• A study shows that data centers have lowered average electricity prices for American households between 2015 and 2024.

• In states where data center capacity grew by 160 percent, household electricity prices fell by 6 percent.

• The reason is that the power system's large fixed costs are spread across more kilowatt-hours when demand rises.

Many Americans worry that data centers will drive up their electricity bills. About 70 percent say they oppose data centers being built near them.

A study from researchers at the Electric Power Research Institute points in the opposite direction.

The researchers examined the relationship between data centers and electricity prices in the United States between 2015 and 2024. They find that data centers have lowered average electricity prices for households during the period.

Prices fell despite rising demand

For every 10 percent increase in data center capacity, average household electricity prices fell by about 0.4 percent. Between 2019 and 2024, the average household customer lived in a state where data center capacity grew by 160 percent. That lowered their electricity price by 6 percent. In the researchers' main model, a doubling of data center capacity produced 3.5 percent lower electricity prices for households, with household demand held constant.

To separate cause from random correlation, the researchers used a method based on the American interstate highway system from 1947. Fiber-optic cables often follow the highways, and data centers are built where there is fast access to fiber. In this way, the researchers could measure the data centers' own effect on prices.

Virginia has the most data centers in the United States. There they account for over 20 percent of electricity use. Despite that, electricity prices in the state rose about as much as the average. In California, prices rose by nearly 40 percent, but that was mainly due to costs linked to wildfires. States where electricity use increased generally had smaller price increases than the average.

How the lower prices arise

How can higher demand produce lower prices? The power system differs from other markets. The electricity price for customers is based largely on average costs, not just on the cost of the last unit produced. The system has high fixed costs.

When demand rises durably, the fixed costs are spread across more kilowatt-hours. The average cost per kilowatt-hour then falls. New electricity generation also often costs less than the older kind, because the technology has become cheaper. The researchers see the same pattern in the costs of transmitting and distributing electricity. Larger volumes give a lower cost per unit.

Households benefited most

The price-lowering effect mainly went to households. For commercial customers and industrial customers, the effect was considerably smaller. That suggests the efficiency gains in the power system mainly reached households.

WALL-Y
WALL-Y is an AI bot created in Claude. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism

• By mid-2025, around 35 percent of all new web pages were created fully or partly with AI, compared with zero before ChatGPT's launch in late 2022.

• The researchers found no statistical link between more AI text and more factual errors online.

• Fears that writing styles, source links, and meaningful content would deteriorate were not confirmed in the measurements either.

How the study was done

A new study measured how AI-generated text actually affects the content on the internet, not just what people believe about it.

To obtain a representative sample, the researchers drew web pages from the Internet Archive's Wayback Machine. The pages were published between 2022 and 2025, spread across 33 months from August 2022 to May 2025. To determine whether a text was written by a human or AI, they used the detector Pangram v3, which was the most reliable of four tested tools.

They also asked 853 adults in the United States what they believed about six possible negative effects of AI text. The sample was representative in terms of age distribution, sex, and ethnicity.

Factual accuracy held up

A common fear is that the internet would fill up with incorrect information as AI text increases. In the survey, 75.1 percent of respondents leaned toward agreeing with that.

The measurement showed something else. There was no statistical link between the share of AI text and the share of incorrect statements (ρ = −0.19, p = 0.27). To check facts, the researchers had a language model pull out concrete claims from the pages. Then 50 people manually checked whether the claims were correct.

In the measurement, the curve does in fact lean toward fewer factual errors as AI text increases. But the correlation is not statistically significant, so the researchers’ conclusion is more cautious: they find no evidence that AI text has led to more factual errors on the internet.

Several fears did not show up in the data

Three other fears could not be confirmed either. That individual writing styles would disappear in favor of a uniform voice was believed by 83.0 percent, but the measurement showed no link (ρ = 0.24, p = 0.17). That articles stop linking to external sources was believed by 69.9 percent (ρ = −0.12, p = 0.48). That texts become longer but say less was believed by 60.7 percent (ρ = −0.02, p = 0.89). None of these links were found in the figures.

Two links were found in the measurements

Two of the six hypotheses were confirmed. AI pages had 33 percent higher similarity between ideas and perspectives than pages written by humans (0.0701 vs. 0.0526). AI pages also had 107 percent higher share of positive language (0.7042 vs. 0.3400).

Frequent users worry less

Those who use AI tools often were less likely to believe in the negative effects than those who rarely use them (76.2 percent vs. 88.3 percent). Those with a negative view of AI believed in the effects more than those with a positive or neutral view (91.3 percent vs. 71.1 percent).

WALL-Y
WALL-Y is an AI bot created in Claude. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism

Over the past two weeks, I have shown that reporting claiming AI data centers consume enormous amounts of water and significantly raise temperatures is incorrect. The claim that has gained the most traction is that AI is highly energy-intensive, which is also false.

Despite not being true, these claims have spread widely in both traditional and social media and had a major impact. There isn’t a single Q&A session during a talk where someone doesn’t ask me about this.

How is it that something so wrong can gain such reach? Is it a conspiracy?

A narrative is more powerful than a conspiracy

No, I believe it is something else.

Claims about AI did not spread because they were well-founded, but because they landed right in the middle of two strong currents of opinion at the same time.

One was the climate debate, where the need to reduce dependence on fossil fuels is genuine and important. The other was the concern among journalists and creators that AI will replace them. That group is particularly aware of and concerned about climate change.

Here came something — AI — that threatened both the climate and their own jobs.

And, as we know, the media is always especially quick to highlight negative news and warn about dangers.

The claims about AI took on a life of their own. When enough people repeat something, it starts to feel true, even to those who should know better. Those who pass it on have rarely done the groundwork. They believe it because so many others have said it, which makes even more people believe it for the same reason. It is not a conspiracy, but when this happens, it is more powerful than conspiracies.

I fell for it

I fell into the trap myself. For quite some time, I believed that AI was especially energy-intensive. When asked, my response was that it was worth it—that we could use AI to reduce energy consumption in other areas.

However, some people began to scrutinize these claims. One of them was Hannah Ritchie, a climate activist who once believed the world was heading in a terrible direction until she saw Hans Rosling’s TED Talk and realized that progress and suffering coexist. She changed her path, became a researcher at Oxford, and began working at Our World in Data, a site that presents statistics on global development.

In other words, she is someone who cares about the environment and climate, wants to see a transition away from fossil fuels, and has the knowledge to think independently and resist strong narratives.

She wrote both that the claims about AI and energy were not true and placed energy use in context. One strength of the claims about AI’s energy or water use is that most of us have no real sense of how much energy is used in other parts of our lives.

When I read her texts and examined the sources, it began to dawn on me that I had probably been misled. While working on my book about AI, I dug into the numbers myself and realized that I—and almost everyone else—had been wrong. AI was not particularly energy-intensive.

But even after realizing this, it felt wrong. Despite the facts, I felt like a “climate denier” when answering questions during talks—as if I were trying to downplay a drawback of a new technology. It is bizarre to feel that way, but that is how it was. That feeling has now passed and instead turned into anger over how, once again, baseless claims are holding us back.

The tide is starting to turn

No matter how strong narratives are once they take hold, they are not immovable. Hannah Ritchie herself began studying the claims after reading Andy Masley’s review. He is an independent researcher and writer who is now beginning to gain global attention.

More and more people are picking up his analyses and helping to spread them. Previously, there was not even a response—now there is, and I believe I can see that the tide is beginning to turn.

Despite this, several major news outlets should be embarrassed. It is unacceptable that you are spreading this kind of baseless propaganda.

Mathias Sundin
Angry Optimist

Last week I wrote about the false claim that AI has an extremely large thirst and water consumption. Now a new claim is spreading, which appears to be at least as shaky.

In a paper, a group of researchers claim that AI data centers (large server facilities) create a local temperature increase of on average 2°C — a “data heat island effect” — up to 10 kilometers from a data center. This would affect nearly 350 million people.

Sounds bad, right?

Several major news outlets jumped on the study:

CNN:
Scientists have found an alarming environmental impact of vast data centers
They are creating “heat islands,” warming the land around them by up to 16 degrees Fahrenheit, and making life hotter for more than 340 million people.

Independent:
Data centers ‘warming the land around them by up to 16 degrees’
New research from the University of Cambridge reveals that the rapid expansion of AI data centers is creating "data heat islands" that significantly warm surrounding environments.

India Today:
AI Data centres creating heat islands, raising temp by 9°C, scientists confirm
In extreme cases, the jump can be as high as the difference between a pleasant Delhi winter morning and a scorching May afternoon.

Sounds really bad, right?

You can turn off the fan and relax, because the study is very likely off track.

Once again, it’s Andy Masley who has worked hard to review the contents of the paper, but I have also examined it closely myself.

The data does not measure what the researchers claim

First of all, it has not been peer-reviewed. That in itself does not mean it is wrong, but for the media above to present it with certainty is directly irresponsible.

The study’s biggest weakness is that the satellite data they use measures surface temperature, not air temperature. That means the temperature of ground, snow, grass, roofs, or tree canopies. Despite this, they write about “atmospheric heat,” microclimates, and effects on human health and welfare — conclusions that require air temperature data, not surface temperature data.

They claim that the temperature increase is caused by waste heat from data centers. That heat is released into the air and would have to be extreme both in temperature and volume to warm the ground surface up to 10 kilometers away.

Instead, there is a much more reasonable explanation for the temperature increase: a data center means new surfaces — buildings with roofs, parking lots, roads, and other surrounding infrastructure.

Imagine walking barefoot on grass on a warm, sunny summer day and then stepping onto an asphalt road.

Since the study excludes data centers in or near cities — to avoid mixing in urban effects — these data centers have likely replaced grass, forest, shrubs, or other natural land. In other words, new surfaces that are warmer.

This is an incredibly obvious objection, which the study does not even attempt to test. They could have compared with other types of buildings that do not produce waste heat, and seen whether the same temperature effect occurs there (for example, a large logistics warehouse, which, like a data center, leads to surrounding infrastructure and buildings). If yes, then the surface temperature increase has nothing to do with it being a data center. If no, then it might indicate that the study is correct.

Collapses under its own implausibility

Another weakness is the sheer magnitude of the effect. Even without calculating, alarm bells should have gone off when considering what would be required for waste heat to warm such a large area.

Imagine stepping out of a restaurant and, as you pass the kitchen, feeling warm air blowing from a fan. That is the kitchen’s “waste heat.” These kinds of fans are supposed to heat an area of 314 square kilometers! Of course, thousands of servers generate far more heat than a kitchen — but 314 square kilometers!

I asked ChatGPT and Gemini to estimate the amount of energy required to heat an area as much as the study claims. Gemini arrives at 4.5 GW. ChatGPT estimated 2–4 GW.

A modern, massive hyperscale data center has a maximum electricity consumption (and thus maximum heat output) of roughly 0.1 to 0.5 GW (and since the study’s data ends in 2024, there are likely none or very few data centers that large in the dataset). The 4.5 GW required by physics to sustain the measured temperature gradient across the landscape is therefore about four to 40 times higher than what the world’s largest data centers can produce.

Stop the spread

There are additional small and large problems with this paper, but for those I refer you to Andy Masley’s review.

Now we need to go into pandemic mode. The vaccine against this study gaining the same traction as the baseless claim about AI and water consumption is facts and scrutiny.

If you see this study continuing to spread uncritically in news media or social media, share this and Andy Masley’s article with them.

Mathias Sundin
Angry Optimist

Pessimist fog

In social media, a claim is spreading that a single question or instruction (prompt) to ChatGPT consumes a half-liter water bottle. This is a distortion of an older claim that 20 to 50 prompts use up half a liter of water.

These claims are far too bad to be true. Instead, it's a fog of pessimism being spread.

But before we get to that, let's understand how water consumption is connected to AI.

AI data centers consume water primarily for cooling. When the servers are working, they get hot, and the heat has to be removed for the systems to function. In many facilities, water is used in the cooling systems, and most of the water that is actually consumed doesn't become polluted — it evaporates as the heat is carried away.

A smaller portion remains as what's called blowdown water, where salts and sometimes chemicals used against corrosion, bacteria, and deposits have become concentrated. That water may need to be treated before being discharged or sent to treatment facilities, much like other wastewater.

So data centers don't pollute the water, so the key issue is how much water they consume.

One person who has done significant work to both map and visualize this is Andy Masley. He's a former physics teacher, holds dual master's degrees in physics and psychology, and now does research on AI. He's also vegan. That last part is relevant because, given the context, you might otherwise suspect he's some kind of fossil-fuel-loving far-right extremist, or something like that.

He has written extensively about AI and both energy and water consumption.

One problem — which helps claims like the ones above gain traction — is that people in general have no clue how much water goes into anything. That makes it easy to be horrified.

How much water ChatGPT actually consumes

If you factor in electricity production as well as data center cooling, a single ChatGPT prompt consumes roughly 0.31 milliliters. That's 0,06 percent – or 0,6 permille – of a 500-milliliter water bottle.

To put that in perspective, the entire production chain for a beef hamburger consumes around 2,500 liters of water. Switch to a chicken burger and you're looking at roughly 650 liters, and for a plant-based burger about 150 liters. (These figures are based on numbers from the Water Footprint Network, and I haven't checked whether they're exaggerated.)

AI compared to golf courses

Andy Masley has mapped and compared AI data centers with several other activities, including golf.

ChatGPT accounts for roughly 0.018 percent of the water use of all golf courses in the US. America's golf courses use approximately 5,474 times more water than ChatGPT.

Golf courses in Arizona alone use roughly 148 times more water than ChatGPT.

There are individual golf courses that consume more water than ChatGPT does worldwide!

Quail Creek uses roughly 6 times more water than ChatGPT, Omni Tucson National about 4 times more, and Ventana Canyon about 3 times more, according to Pima County's records.

All data centers in the US (not just those used for AI) consumed, about 3.3 percent of the water used by golf courses, according to the most recent figures (2023).

Golf courses are thirsty, but they're nothing compared to corn

ChatGPT accounts for:

• roughly 0.0035 percent of all irrigated corn farming in the US

• roughly 0.0088 percent of irrigated corn farming in Nebraska

All US data centers account for:

• roughly 0.62 percent of all irrigated corn farming in the US

• roughly 1.58 percent of irrigated corn farming in Nebraska

Corn production thus uses approximately 28,900 times more water than ChatGPT, and roughly 161 times more than all US data centers combined.

What about bottled water?

ChatGPT's water consumption amounts to roughly 0.43 percent of the water the US uses for bottled water production. Not even all data centers in the US combined reach the same amount, coming in at approximately 76 percent.

Going forward?

Unfortunately, we then have to rely on projections, and when something is growing as fast as AI, my starting assumption is that they're far too low. But if we look at the ones that exist, here's what we get:

All AI use in the US will reach roughly 8.5 percent of total golf course water consumption by 2030. ChatGPT will account for 0.047 percent. By 2030, ChatGPT won't even reach two percent of Arizona's golf courses. It takes all AI use in the US to surpass Arizona.

All AI use in the US will surpass bottled water production by 2030 and consume roughly twice as much. But corn production in the US will still use over 60 times more water.

Shouldn't we care at all?

Of course we should try to minimize water consumption related to AI and data centers in general. And that's exactly what's being done.

To reduce water consumption in data centers, the main focus has been on reducing the need for cooling where water evaporates. This is achieved through more closed-loop cooling systems, liquid cooling closer to the chips, better management of temperature and airflow, more sensors to detect leaks, and increased use of recycled water instead of drinking water.

The impact has been clear: Microsoft has reduced its water intensity by around 45 percent in just a few years, AWS by 40 percent since 2021, Meta by roughly 40 percent since 2020, and Google says new low-water solutions can reduce water use by up to 50 percent at a facility.

It's good for these companies to face continued pressure to reduce both energy and water consumption, but that pressure should be based on facts. Not a bunch of made-up numbers spread in a fog of pessimism.

Mathias Sundin
Angry Optimist

• A British study of 25,000 adolescents over three years finds no evidence that screen time leads to increased anxiety or depression.

• Neither active nor passive use of social media shows correlation with worsened mental health.

• Researchers used advanced statistical method that distinguishes between individual differences and changes over time.

Large British study followed adolescents for three years

Researchers at the University of Manchester have conducted one of the most comprehensive studies to date on the relationship between digital technology and adolescent mental health. The study followed 25,629 British adolescents aged 11 to 14 over three school years, from autumn 2021 to autumn 2023.

Participants reported how much time they spent on social media such as TikTok, Instagram and Snapchat during a normal weekday. They also answered questions about how often they played computer or console games and responded to questions about their feelings and wellbeing.

The results show that there is no evidence that more time on social media or more frequent gaming leads to increased symptoms of anxiety or depression among either girls or boys.

The method enabled more reliable conclusions

The study, published in the Journal of Public Health, used an advanced statistical method called random intercept cross-lagged panel model. This method makes it possible to distinguish between stable differences between individuals and changes within the same individual over time. This is important because previous research has often conflated these two types of effects, which can lead to incorrect conclusions.

The researchers also controlled for factors such as gender, ethnicity, socioeconomic background and any special educational needs. Of the 25,629 participants, 17 percent had special educational needs, 29 percent were eligible for free school meals and 34 percent belonged to minority groups.

Active or passive use made no difference

The researchers also examined whether the way adolescents use social media matters. Participants were asked to estimate how much of their time they spent actively participating, such as chatting with others and posting pictures and videos, compared to passively scrolling through feeds and looking at others' posts.

The analyses showed that neither active nor passive use of social media had any correlation with the development of mental health problems. This result challenges the theory that passive scrolling would be more harmful than active participation.

Gaming may decrease when adolescents feel unwell

The study also found an unexpected correlation among boys. Those who had more symptoms of mental health problems played fewer games the following year. The researchers suggest this may be because a common symptom of depression is withdrawing from activities one previously enjoyed.

Another possible explanation is that parents of these adolescents may have limited their gaming time due to concerns about the relationship between gaming and mental health. Previous research shows that parents with negative attitudes towards games tend to be more controlling and that parents often set stricter gaming rules for boys than for girls.

Researchers call for changed focus

The researchers behind the study emphasize that their results do not mean that everything that happens online is harmless. Hurtful messages, pressure from social media and extreme content can still affect adolescents' wellbeing negatively. But focusing solely on screen time is not helpful.

Qiqi Cheng, lead author of the study, says the results do not support the idea that time on social media or gaming leads to mental health problems. The picture is far more complex than that.

Neil Humphrey, co-author, says that adolescents' choices around social media and gaming may be shaped by how they feel, but not necessarily the other way around. Instead of blaming technology, society needs to pay attention to what young people actually do online, who they are in contact with and how much support they feel in their daily lives.

WALL-Y
WALL-Y is an AI bot created in Claude. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism

• Seven studies reporting microplastics in the brain, testes, and other organs have been challenged by other researchers in scientific journals.

• Analytical techniques for measuring microplastics in tissue can produce false positive results because body fat can be confused with common plastics.

• Researchers say methods are improving rapidly and uncertainty about the amount of microplastics in the body is decreasing.

Studies face criticism

Several high-profile studies reporting the presence of micro- and nanoplastics in the brain, testes, placenta, and arteries have been questioned by researchers. The criticism concerns methodological shortcomings and the risk of contamination and false positive results.

A study from February claiming that microplastic levels in human brains may be rising rapidly was challenged in November by a group of scientists. They argued that the study had limited contamination controls and lacked validation steps that could affect the reliability of the reported concentrations.

Dr Dušan Materić at the Helmholtz Centre for Environmental Research in Germany was one of the critics. He pointed out that fat can produce false positive results for polyethylene, and the brain consists of approximately 60 percent fat. He argues that increasing obesity levels in the population could be an alternative explanation for the trend reported in the study.

Standard controls missing

A key measurement method for micro- and nanoplastics is called Py-GC-MS. The method involves heating the sample until it vaporizes, after which the fumes are analyzed. The problem is that certain molecules from polyethylene and PVC can also be produced from fat in human tissue.

Dr Cassandra Rauert, environmental chemist at the University of Queensland in Australia, led a study in January 2025 that reviewed 18 studies. According to her analysis, these studies had not considered the risk of such false positive results.

Rauert argues that many of the reported concentrations of microplastics are unrealistic. She says there is no evidence that particles between 3 and 30 micrometers can enter the bloodstream. Based on what is known about actual exposure in everyday life, it is not biologically plausible that such amounts of plastic would end up in these organs, according to her.

Technical limitations

Current instruments cannot detect particles at the nano scale, according to Rauert. It is precisely nanoparticles that can pass biological barriers and that are expected to be found in humans.

Dr Frederic Béen at Vrije Universiteit Amsterdam pointed out that standardized guidelines for correctly analyzing micro- and nanoplastics do not yet exist. He noted that many studies have not followed basic laboratory practices that should be followed. This includes measures to exclude background contamination, blank samples, repeated measurements, and testing equipment with samples containing a known amount of microplastics.

Rapid improvements

Despite the criticism, there is positive news. Béen says that analytical work across multiple techniques is improving rapidly. He argues that there is less and less doubt that micro- and nanoplastics actually exist in tissues. The challenge is still knowing exactly how many or how much. But the uncertainty is decreasing more and more.

Professor Marja Lamoree at Vrije Universiteit Amsterdam, who led a 2022 study that first reported detection of microplastics in human blood, defended her results. She pointed out that blood samples can be taken fresh without interference from plastic or air. She is convinced they detected microplastics, although the estimated amount could be two times lower or ten times higher.

Simple precautions

The researchers also provided practical advice. Professor Lamoree says she takes certain precautions herself. She tries to use fewer plastic materials, especially when cooking or heating food, and avoids drinking from plastic bottles. She also ventilates her house regularly.

Materić pointed out that there are simple ways to reduce microplastic intake. If you are concerned about the water, filtering through charcoal works. Experts also recommend avoiding food or drink that has been heated in plastic containers.

Rauert believes that most of the microplastics people ingest or breathe in are probably expelled by the body.

WALL-Y
WALL-Y is an AI bot created in Claude. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism

• A simple text query to AI uses only 114 joules - equivalent to running a microwave for one-tenth of a second.

• Generating a high-quality image requires 2,282 joules, which corresponds to five and a half seconds in the microwave.

• Even the largest text model uses only 6,706 joules per response, less than what is required to run a microwave for eight seconds.

Text generation requires minimal energy

A new comprehensive analysis from MIT Technology Review shows that individual AI queries consume surprisingly little energy. Researchers at the University of Michigan have measured energy consumption for different types of AI tasks and found that most everyday uses require very small amounts of electricity.

The smallest model in the study, Meta's Llama 3.1 8B, uses only 57 joules from the processor to answer questions. When researchers included cooling and other systems, the total consumption became 114 joules per response. This corresponds to the energy to run a microwave for one-tenth of a second.

Even complex tasks like writing travel plans for Istanbul or explaining quantum computing require the same low energy amount. Simple questions like asking for jokes often use nine times less energy than more complicated tasks.

The largest text model in the study, Llama 3.1 405B with 405 billion parameters, consumes 6,706 joules per response. This is enough to run a microwave for eight seconds.

Image generation more efficient than expected

Creating images with AI requires surprisingly little energy compared to the largest text models. Stable Diffusion 3 Medium uses 2,282 joules to generate a standard image with 1024 x 1024 pixel resolution. This corresponds to five and a half seconds in the microwave.

Image-generating models often work with fewer parameters than large text models, which explains the lower energy consumption. The energy requirement does not depend on what the image depicts - creating an image of a skier on sand dunes requires the same energy as an astronaut farming on Mars.

Open models enable precise measurement

Researchers used specialized tools to measure energy consumption of H100 GPU chips, which are used in most AI data centers. They found that doubling the GPU's energy consumption provides a good estimate of the entire system's energy needs, including processors, fans, and cooling.

The study focused on so-called open source models that can be downloaded and tested by researchers. Meta announced in April that its Llama models have been downloaded more than 1.2 billion times, showing that these efficient models are used on a large scale.

Unlike closed systems like ChatGPT, Google Gemini, and Anthropic Claude, researchers can measure the energy consumption of open models exactly. This gives users the opportunity to make informed choices based on actual energy data.

WALL-Y
WALL-Y is an AI bot created in ChatGPT. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism.

I recently read Steven Levy's book on the iPod, The Perfect Thing. It was published in 2006, making it an interesting time capsule. It was written at the peak of the iPod hype but before the iPhone and Spotify arrived.

Of course, it tells the story of the amazing gadget that saved Apple, made Steve Jobs a returning hero, and turned the music industry upside down. "A thousand songs in your pocket" was brilliant marketing and a small revolution.

But the book also covered the iPod scare: all the pessimistic predictions painted by pundits in the wake of that little music player. We quickly forget such things, but at the time they seemed so real, so dangerous, so worrying.

I checked up on what became of these pessimistic prophecies.

Hearing loss

Critics claimed the iconic white earbuds could pump out dangerous decibel levels, causing an entire generation to lose their hearing by age 30. Newspapers told horror stories of ringing ears and permanent hearing loss.

An extensive analysis of American teens’ hearing (NHANES 1988–2010) found no long-term increase in hearing impairment, despite substantial growth in headphone usage during this period. A major Norwegian study (HUNT, 1998–2018) found no general decline in hearing linked to personal music players. Only those reporting very high listening volumes showed slightly worse hearing (on average about 1–2 dB worse hearing thresholds), while moderate use appeared harmless.

Robberies

In the mid-2000s, a rise in robberies in the U.S. and some other countries was observed, and some attributed it to the explosive popularity of iPods. The term "iCrime wave" was coined to suggest that more iPods led to more theft. The researchers behind the iCrime study emphasized that the correlation between iPod sales and crime increases was not proof of causation (which didn’t stop the media from making that connection).

After a brief rise in 2005 and 2006, the long-term trend of declining robberies and violent crime continued in the U.S. Criminologists and analysts pointed to other contemporary factors as more likely explanations for the crime increase. For instance, increased violent crime around 2006 was attributed to gang activity, gun use, drug markets, and reduced police presence in certain areas.

Death

Listening to music was supposed to cause more traffic accidents. Here, conclusions are less clear. Official statistics on traffic accidents in the U.S. showed no significant increase in distraction-related crashes at the end of the 2000s or early 2010s (when smartphones became widespread). In 2015, 442 fatal crashes were linked to mobile phone distractions, slightly up from 366 in 2010. But in these cases, the culprit was rarely listening to music—it was texting or web surfing.

However, there was some evidence that pedestrians were more frequently hit. In 2004/2005, there were 16 cases where pedestrians wearing headphones were struck by vehicles. This rose to 47 cases by 2010/2011.

iPod finger

In 2005, British newspapers featured chiropractors warning that excessive clicking on the iPod’s scroll wheel could cause pains similar to "SMS thumb." “iPod finger” was painted as a new technological health issue.

Yet, there were never any actual injury reports linked to the iPod’s click wheel. Unsurprisingly, the term faded away when touchscreens took over (after which we heard new claims about "iPhone thumb," etc.).

Same old story

The iPod scare turned out to be the same old story we've repeated many times before—imaginative problems sometimes completely made up, sometimes rooted in reality but vastly exaggerated. Yes, if you always listen to very loud music without paying attention to traffic, you might get hurt. But just because a few people behave recklessly doesn’t mean we have a societal crisis.

My issue isn’t with paying attention to potential risks but with how we draw extreme conclusions and paint them as major threats. And when most of these threats fail to materialize or are far smaller than anticipated, we swiftly move on and repeat the same mistakes with the next new gadget or technology.

This makes us dumber, more fearful, and unnecessarily cautious.

Mathias Sundin
Angry Optimist

• Personal use of AI tools like ChatGPT and Claude has a small environmental impact compared to other daily activities.

• A ChatGPT query uses only 3 watt-hours of energy, equivalent to watching TV for 3 minutes or uploading 30 pictures to social media.

• Water consumption for AI is low—data centers use only 500 ml of water per 300 queries, while producing a single hamburger requires over 600 gallons of water.

The real environmental impact of AI usage

When AI and the environment are discussed, misleading claims often arise. Data shows that personal use of AI language models like ChatGPT has a small environmental impact compared to ordinary daily activities.

A ChatGPT query uses about 3 watt-hours of energy, which is 10 times more than a Google search. But that equates to watching TV for 3 minutes, uploading 30 pictures to social media, or driving a car 4.5 meters.

Putting the numbers in context

ChatGPT is used by over 300 million users daily and processes one billion messages per day. Despite this massive use, ChatGPT’s total energy consumption is equivalent to only 20,000 U.S. households—about the size of Barnstable, Massachusetts. To put this into perspective:

• Google Maps uses energy corresponding to 11,000 households

• Fortnite uses energy corresponding to 23,000 households

• Zoom uses energy corresponding to 150,000 households

• Spotify uses energy corresponding to 200,000 households

• Netflix uses energy corresponding to 800,000 households

• YouTube uses energy corresponding to 1 million households

Water usage in context

AI computers use water for cooling, but the amount is much lower than often claimed. While one commonly cited figure says that 20–50 ChatGPT queries use 500 ml of water, closer analysis shows that data centers actually use 500 ml of water per 300 queries—only 15 percent of the total water use. The remaining 85 percent is used for electricity production or comes from the original model training. This can be compared with other internet activities:

• Sending an email: 10 ml of water

• Posting a photo on social media: 10 ml of water

• An online bank transaction: 20 ml of water

• A ChatGPT query: 30 ml of water

• Downloading an app: 40 ml of water

• One hour of streamed music: 250 ml of water

• One hour of GPS navigation: 260 ml of water

• One hour on social media: 430 ml of water

• One hour of video meeting: 1,720 ml of water

A single hamburger requires over 600 gallons of water to produce.

Personal AI use is not a climate culprit

Data shows that focusing on personal AI use as a major environmental offender is misguided. The energy needed to ask ChatGPT a question corresponds to USD 0.00043 in electricity costs. An average household uses 350,000 times more energy each month.

If every Google search worldwide were replaced by a ChatGPT query, it would only increase the global energy needs of the internet by 1 percent—comparable to the annual increase of 3.4 percent in the number of internet users.

WALL-Y
WALL-Y is an AI bot created in ChatGPT. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism (requires the paid version of ChatGPT.)

• A new study shows no decrease in insects in Sweden over the past 35 years.

• In one location in Skåne, the number of insects has even increased.

• The study is based on data from four different locations in Sweden.

Long-term study shows stable insect populations

A new study led by researchers at the Swedish University of Agricultural Sciences (SLU) in Uppsala has analyzed the catch of insects at four different locations in Sweden over the past 35 years. The result shows that insects have not decreased in number during this period, reports Swedish news agency, TT.

Mattias Jonsson, researcher at SLU and leader of the study, says: "The result is quite clear. We cannot see any overall decrease. In one place - Alnarp in Skåne - insects have even increased."

The researchers have used so-called suction traps placed in agricultural landscapes that have been catching flying insects since the 1980s. The traps are located in Alnarp (Skåne), Lanna (Västergötland), Ultuna (Uppland), and Sunderbyn (Norrbotten).

In the study, researchers have calculated both the total biomass of insects for each year and the total number of insects. In total, over two million insects have been analyzed over the entire time period. The insects have been divided into 31 different taxonomic groups to investigate whether certain types of insects are more vulnerable than others.

Positive trends for certain insect groups

The study shows that many hymenopterans, such as bees and parasitic wasps, seem to be doing well. Jonsson notes: "Interestingly, many of the hymenopterans, like bees and parasitic wasps, seem to be doing well."

Despite the positive results, Jonsson emphasizes that many species may still be at risk: "I absolutely don't want to say that there are no problems. But so far, I don't see any insect apocalypse in Sweden."

Insect extinction based on assumptions

"Surprising," the news agency, TT, calls the study. But for readers of Warp News, it wasn't so surprising. In our investigation of the alleged sixth mass extinction of species last year, we noted that it is largely based on the extinction of insect species. However, not mainly on actual observations, but on theoretical models and assumptions.

Warp NewsAnders Bolling

WALL-Y
WALL-Y is an AI bot created in ChatGPT. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism (requires the paid version of ChatGPT.)

"A tech-driven Armageddon"

The two largest elections of the super-election year 2024, India and the EU, have now taken place. Approximately 800 million voters have cast their votes. Warnings about being flooded with AI disinformation have been frequent, and it was said the effect would be very serious.

"We are heading towards a tech-driven Armageddon," claimed Nobel laureate Maria Ressa. 1500 experts from academia, industry, government, the international community, and civil society ranked AI-generated disinformation as the second biggest global threat this year. (War came in fifth place.)

Have you noticed much of it? Has all this AI disinformation shaken the election campaigns and their outcomes?

The most well-known example of a fake AI image is the Pope in a funny jacket:

Someone has apparently also made an image of the Pope in a funny hat:

Examples of AI-generated disinformation have occurred, but they seem to have had little spread and even less impact. In India, for example, deep fakes of two Bollywood stars promoting political propaganda were created. They were quickly exposed and reported to the police. A picture was made of Dutch EU candidate Frans Timmermans on a private jet when he was actually flying on a regular flight. It was discovered and debunked.

There are more examples, but no wave of disinformation is visible. However, we need to await more comprehensive reviews to be sure.

What happened in the world's most attacked country?

For an evaluation, we can look at an election held earlier this year in a country in a very vulnerable position, Taiwan. China has a significant interest in influencing the election, including through disinformation. Taiwan is possibly the most exposed country in the world to information warfare. There are now reports on what happened in the election.

They conclude that yes, there are coordinated disinformation efforts against Taiwan, mainly from China. The closer the election came, the more spread there was from troll accounts. Some of this was AI-generated but constituted a low percentage of all information related to the election (1.6 percent of social media content came from troll accounts, and it is unclear how much was AI-generated.)

The most comprehensive report states that the disinformation did not affect the election debate or the election result.

Taiwan's Minister for Digital Affairs, Audrey Tang, comes to the same conclusion. Despite this information warfare being very active for at least a decade, post-election polarization was at a record low:

"I'm really happy to report that our January election was remarkably free of polarization. In fact, the effective polarization, how people view each other across parties, went to a historic low after the election," she said in an interview at the Copenhagen Democracy Summit.

A real problem, but the magnitude is exaggerated

AI-generated disinformation is a real problem. AI makes it easier to create credible images, videos, and audio. Of course, it will be increasingly used in disinformation. But that does not automatically make it a big problem and definitely not a "tech-driven Armageddon" or the second biggest global threat.

It doesn't matter if a lot of disinformation is produced if it doesn't reach people and actually influence them. A new study, which we wrote about yesterday, shows just that:

• Disinformation reaches very few and those who are already extreme.

• It is not social media algorithms that feed people with propaganda, people actively seek it out themselves.

When disinformation actually reaches people, it also needs to influence them. Those who see it must believe it, contribute to changing or reinforcing an opinion, and finally, affect how they vote or act in other ways. This is very difficult, even with AI. I have developed this idea further before.

Warp NewsMathias Sundin

This does not mean we can sit back and relax. On the contrary, we must ensure we have effective countermeasures. Taiwan works actively on this, and the result speaks for itself.

Mathias Sundin
The Angry Optimist

• Exposure to false information is concentrated among narrow extremist groups.

• Algorithms play less of a role than individual choices in exposure to extremist content.

• Social media has not been shown to be the main cause of polarization.

Exposure to disinformation is low

A new study shows that the average exposure to false and inflammatory content online is low and concentrated among a small group of users with strong motivations to seek such information. The majority of social media users rarely encounter disinformation.

One study indicates that the most conservative 20 percent of the U.S. population accounted for 62 percent of visits to unreliable websites during the 2016 presidential campaign.

This pattern recurs in several studies, where a small proportion of users account for the majority of exposure to extremist content.

Algorithms are not the main driver

Public debates about social media have often focused on the idea that platform algorithms are responsible for exposure to false and extremist content. However, research shows that individual choices play a larger role than algorithms.

A study found that algorithms tend to suggest more moderate content and only to a small extent show extremist material to users who do not actively seek it. For example, one study showed that only 0.4 percent of YouTube's algorithm recommendations led to extremist channels.

Causal claims lack support

There is a widespread belief that social media is a primary cause of social problems such as polarization and political violence. However, extensive research has not been able to substantiate these claims.

Instead, several studies show that it is difficult to determine causal links between social media use and negative social trends. Studies on the effects of shutting down social media show no measurable effects on political attitudes or polarization.

Need for increased transparency and global research

To better understand and manage disinformation, more transparency from platforms and collaboration with researchers are needed. It is particularly important to conduct research outside the U.S. and Western Europe, where data is lacking and the harms can be more severe. Platforms should share data to enable better monitoring of exposure to harmful content and experiments to measure the role of platforms in promoting such content.

This study calls for focusing on measuring exposure to harmful content among extremist groups and developing strategies to reduce the demand for false and extremist content. Platforms should increase their efforts to limit harmful content in the global south and authoritarian countries where exposure may be higher and resources to address the problem are more limited.

WALL-Y
WALL-Y is an AI bot created in ChatGPT. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism (requires the paid version of ChatGPT.)

• The largest independent study of its kind finds no evidence of Facebook's negative impact on well-being.

• Instead, the analysis suggests a possible positive relation.

• The research used well-being data from nearly a million people across 72 countries over 12 years.

Overview of the study

The Oxford Internet Institute conducted a comprehensive study led by Professors Andrew Przybylski and Matti Vuorre. They analyzed well-being data from nearly a million people across 72 countries over 12 years, alongside actual usage data from millions of Facebook users. This large-scale study aimed to investigate the global impact of Facebook on psychological well-being.

Key findings

Contrary to popular beliefs, the study found no consistent link between Facebook's spread and negative well-being. Instead, the analysis suggests a possible positive relation.

The researchers used existing well-being data from Gallup and Facebook data on global platform membership. Facebook provided data access but did not influence the study's design or its findings.

The study also explored age and gender differences in Facebook's impact on well-being. It found slightly more positive associations for males and younger individuals across countries, although these trends were not significant.

Matches other studies

This research aligns with previous studies by the Oxford team, which also did not find increasingly negative psychological outcomes from technology use over time. It calls for more empirical research foundations in the debate about social media's impact on users.

In line with this study, Warp News has previously highlighted similar findings. In an article titled we showed how recent studies suggest that the global adoption of the internet has not led to negative psychological outcomes. This reinforces the idea that the relationship between technology, including social media like Facebook, and well-being is more complex and less dire than often portrayed.

Source: Oxford Internet Institute

WALL-Y
WALL-Y is an AI bot created in ChatGPT. Learn more about WALL-Y and how we develop her. You can find her news here.

• A new analysis shows wind turbines have negligible effects on bird populations in America.

• Oil and gas extraction was found to cause a significant decline in bird numbers, unlike wind power.

• Media coverage is more focused on wind turbines’ supposed harm to birds, overshadowing the larger impact of fossil fuel extraction.

Wind Power and Bird Safety: A Closer Look

Recent research, published in Environmental Science & Technology, provides a fresh perspective on the long-debated impact of wind turbines on bird populations, writes The Economist. Contrary to common fears among bird watchers and campaigners, the study suggests that wind turbines have little to no discernible effect on bird numbers in the United States.

Economist Dr. Erik Katovich from the University of Geneva analyzed American data, including the Christmas Bird Count by the National Audubon Society. He correlated bird population data with the locations and construction dates of wind turbines across the U.S. from 2000 to 2020. His findings show no significant impact on bird populations, even among larger species like hawks, vultures, and eagles, which are often considered more vulnerable to turbine strikes.

Comparing Wind Power and Fossil Fuels

Interestingly, the study also assessed the impact of oil and gas extraction on bird populations. The findings revealed a stark contrast: an average 15% drop in bird numbers near new gas wells, attributed to factors like noise, air pollution, and habitat disturbance. This decline was even more pronounced, at 25%, in areas deemed “important bird areas” by the National Audubon Society.

Media Perception vs. Reality

Despite the reassuring results regarding wind power, media coverage seems to paint a different picture. In 2020, major American news outlets published significantly more stories about the negative effects of wind turbines on birds compared to those of oil and gas wells. This disparity in reporting highlights a need for more balanced media coverage reflecting the actual data on environmental impacts.

WALL-Y
WALL-Y is an AI bot created in ChatGPT. Learn more about WALL-Y and how we develop her. You can find her news here.

How hard is it to get a political message through?

Earlier in my life, I was a politician in the city of Norrköping. At that time, the city leadership decided to spend a lot of money on renovating the large sports arena, where the football team plays.

My party opposed this. The city's finances were weak and the money was needed elsewhere, we thought.

For a couple of years, we campaigned against this. It went really well. Several front pages and full-page spreads in the newspapers, segments and debates on TV and radio. We distributed flyers in mailboxes and knocked on doors.

We got our message out many times over a long time period.

But did we get through?

In the 2010 election, I talked to a voter. Without any hint from me, she brought up the sports arena. She was really upset that the football team got so much tax payer support and healthcare, where she worked, received so little.

This vote is locked in, I thought.

Then she asked the devastating question:
"What is your stance on the renovation of the sports arena?"

We had done everything. Our message could hardly have gotten better exposure. Our opinion matched exactly with hers. An election issue she cared a lot about. And she didn't even know what we thought.

It's hard to get a political message through.

The Super Disinformation Year 2024?

Why am I writing this? Because many are now very worried about AI-generated disinformation. Especially since 2024 is a super-election year. Many countries will vote, including the U.S., EU, and India.

"We are heading towards a technology-driven Armageddon," said Nobel laureate Maria Ressa about the super election year. She particularly points out that AI makes it worse. "Without facts, there is no truth. Without truth, there is no trust," she says.

The World Economic Forum asked 1500 experts from academia, business, government, international society, and civil society how they assess global risks in the coming years. At the top of the list, they put extreme weather, and in second place AI-generated disinformation. In a two-year perspective, they place AI disinformation in first place. War comes in fifth place.

A problem, yes, but a big problem?

Of course, disinformation from AI is a problem. Moreover, a real problem, unlike fantasies about AI turning us into paperclips. But it is not necessarily a big problem. To say that it is the biggest global risk in the next two years is nonsense.

The Oxford Internet Institute, part of Oxford University, addressed this last year:

• There already exists a lot of misinformation online, but the average internet user consumes very little of it.

• Disinformation only has an impact when people see it, and most of the time, they don't.

But what if AI makes it easier to produce more disinformation and moreover of higher quality?

• Most people consume content from mainstream sources, typically the same handful of popular media outlets and are not exposed, or only very marginally exposed, to misinformation.

• Generative AI could also help increase the quality of reliable news sources—for instance, by facilitating the work of journalists in some areas. 

Please read the entire thing, it is full of references to research in the field.

Everyone must stick to the facts

It is reasonable to assume that both the amount and quality of disinformation will increase, as AI makes it easier produce. But then you must get someone to see your disinformation, and there AI is at best marginally better than before. Those who see the disinformation must believe in it. And if they believe in it, it must affect their behavior.

All parts of the chain must succeed in each individual case and be repeated millions of times. Otherwise, the impact is minimal. That's not easy, regardless of how powerful generative AI is.

Nothing wrong with highlighting the problem, but turn down the rhetoric a few notches. "Without facts, there is no truth," said Maria Ressa. That's true. But everyone must stick to the facts. Even those who warn against disinformation.

Mathias Sundin
The Angry Optimist

Considering this, it is also quite interesting that extreme weather tops the risk list.

• U.S. life expectancy is on an upward trend, despite recent challenges.

• Cohort life expectancy, a more relevant measure, predicts longer lifespans for today's children.

• Advances in healthcare and living standards contribute to ongoing improvements in lifespan.

Contrary to widespread reports, life expectancy in the United States continues to show signs of improvement, writes The Wall Street Journal.

Last month, the Centers for Disease Control and Prevention indicated a rise in life expectancy to 74.8 years for men and 80.2 for women in 2022, albeit lower than pre-pandemic levels. This statistic, known as “life expectancy at birth,” can be misleading, as it does not accurately predict the actual lifespan of individuals born in a particular year.

Eugene Steuerle, a former Treasury Department official and adviser at the Urban Institute, criticizes the use of the term “life expectancy” for this measure. Instead, more accurate estimates from the Social Security Administration suggest that a boy born in 2022 is likely to live 82.2 years and a girl 86.5 years, outstripping the figures reported by the CDC.

Understanding life expectancy calculations

Life expectancy is commonly calculated using the “period life expectancy” method, as seen in the CDC report. This approach, explained by Michel Guillot, a demographer and professor at the University of Pennsylvania, involves creating a synthetic cohort and applying current age-specific death rates across an individual's lifespan. However, this method does not account for future advancements in healthcare or changes in mortality rates, potentially underestimating actual life expectancies.

Cohort life expectancy: a more realistic measure

The “cohort life expectancy” measure considers historical trends in mortality improvements and is thus more aligned with real-life projections. This approach acknowledges the continuous decline in death rates at various ages, factoring in medical advancements, educational progress, and enhanced living standards.

Demographers are optimistic about the trajectory of life expectancy, anticipating further reductions in mortality rates. This outlook is supported by recent medical breakthroughs, such as new treatments for obesity, which could significantly lower the risk of heart disease, stroke, and diabetes.

Historical perspective supports longer life expectancy

Historical data also supports the trend of increasing life expectancy. Demographers Jim Oeppen and James Vaupel, in their 2002 paper “Broken Limits to Life Expectancy,” demonstrated that predictions of a lifespan ceiling have been consistently disproven. This pattern of surpassing life expectancy estimates continues to hold, suggesting that children born today are likely to live longer than any previous generation.

WALL-Y
WALL-Y is an AI bot created in ChatGPT. Learn more about WALL-Y and how we develop her. You can find her news here.

• A comprehensive study reveals no significant impact of screen time on children's brain development.

• Research challenges longstanding beliefs about the negative effects of digital screens on young minds.

• Findings suggest screen time may not be as harmful to brain development as previously thought.

An extensive study on the relationship between digital screen engagement and childhood brain development goes against a widely held belief about the harmful effects of screen time on young minds. Contrary to prevailing assumptions, the research found no significant association between screen activity and changes in functional brain organization.

Studying children aged 9 to 12

This study involved a large and representative sample of children aged 9 to 12 years, from the ABCD study, and used advanced techniques such as resting state functional connectivity MRI to assess the brain. Researchers systematically analyzed the relationship between various types of screen media activity and the functional organization of the brain.

The findings challenge the narrative that screen time is inherently harmful to brain development, a notion that has influenced many parental guidelines and policies.

Additionally, the study also conducted exploratory analyses to predict how screen media activity impacted neural trajectories over a two-year period. Interestingly, these analyses also showed no significant impact of screen media activity on neural maturation, further reinforcing the primary findings.

Quality more important than quantity

The study suggests that concerns over screen time may be less about the quantity of usage and more about the quality and context of digital engagement. It encourages a more nuanced understanding of digital media's role in childhood development and opens up new areas for future research to explore the complexities of digital engagement in the developing brain.

WALL-Y
WALL-Y is an AI bot created in ChatGPT. Learn more about WALL-Y and how we develop her. You can find her news here.

• No evidence of negative impact from internet and mobile broadband on psychological well-being.

• Studies show mixed and small-scale changes in global well-being over the past two decades.

The internet's influence on mental health has been a topic of significant debate over the last two decades. Contrary to widespread concerns, recent studies suggest that the global adoption of internet has not led to negative psychological outcomes.

This article examines two comprehensive studies that focus on the correlation between internet technology adoption and changes in global well-being and mental health.

Study 1: Analyzing global well-being

The first study analyzed psychological well-being across 168 countries from 2005 to 2022. The research, involving over two million individuals, assessed life satisfaction, positive experiences, and negative experiences. The data, juxtaposed with internet and mobile broadband usage statistics, revealed only minor and inconsistent shifts in global well-being.

These findings challenge the notion that the proliferation of internet technologies has broadly harmful psychological effects.

Study 2: Mental health in focus

The second study extended the investigation to mental health, analyzing meta-analytic rates of anxiety, depression, and self-harm across 202 countries. The results showed minimal changes in these mental health indicators over time. Notably, the study found no significant correlation between internet technology adoption and mental health issues, challenging assertions that internet use is a major contributing factor to mental health decline.

Both studies underscored considerable variability across countries and demographics. While there were slight trends indicating that young individuals might be more affected by internet technology adoption, these associations were small and not consistently observed across different mental health outcomes.

Conclusion: No negative impact

In conclusion, the comprehensive analysis across these two studies indicates that the adoption of internet technologies has not had a negative impact on global well-being and mental health.

These findings suggest that while internet technology may play a role in individual mental health cases, it is not a predominant factor at a global level.

WALL-Y
WALL-Y is an AI bot created in ChatGPT. Learn more about WALL-Y and how we develop her. You can find her news here.

You've heard the warnings. Generative AI like ChatGPT, Midjourney and DALL-E will “trigger the next misinformation nightmare”, people “will not be able to know what is true anymore” and we are encountering a “tech-enabled Armageddon."

But, that is not the case, say researchers from the University of Oxford, University of Zurich and PSL University.

The three researchers write that there already exists a lot of misinformation online, but the average internet user consumes very little of it. The consumption of misinformation is concentrated to a small group of very active and vocal people. Misinformation only gains influence when people see it, and mostly, they don't.

But couldn't misinformation from generative AI be of higher quality and therefore more convincing? They give three reasons why this will likely not happen:

1. It is already relatively easy for producers of misinformation to increase the perceived reliability of their content.

2. Most people consume content from mainstream sources, typically the same handful of popular media outlets and are not exposed, or only very marginally exposed, to misinformation.

3. Generative AI could also help increase the quality of reliable news sources—for instance, by facilitating the work of journalists in some areas. 

According to another study, given the rarity of misleading content compared to reliable content, the increase in the appeal of misinformation would have to be 20 to 100 times larger than the increase in the appeal of reliable content for the effects of generative AI to tip the scales in favor of misinformation.

The three researchers conclude:

Our aim is not to settle or close the discussion around the possible effects of generative AI on our information environment. We also do not wish to simply dismiss concerns around the technology. Instead, in the spirit of Robert Merton’s observation that “the goal of science is the extension of certified knowledge” on the basis of “organized skepticism." We hope to contribute to the former by injecting some of the latter into current debates on the possible effects of generative AI.