Kurtis' Blog · What's Wrong With AI?

What's Wrong With AI? The opinions expressed in this blog post are mine and mine alone, and do not reflect the opinions or policies of my employer. I recently started working at a software company, and I was surprised at how ubiquitous AI was. Nearly everyone uses it, but nobody seems to be aware of the problems it causes. So, I've decided to write this blog post, and discuss the various issues with AI as it is currently, ranging from verifiable problems that exist today (with sources!), to the personal opinions and thoughts of myself and others. A few disclaimers before we start: First, when I use the term "AI", I am referring specifically to LLMs, image generators, or other forms of machine learning algorithms that are referred to as "AI" in common parlance. I am not referring to Artificial Intelligence as a concept, Artificial General Intelligence (AGI), or machine learning in general. Second, I am not strictly against AI. In fact, I think it is an incredibly useful tool in certain contexts. My stance is briefly summed up by saying, "I am not against AI, I am against the business practices of the companies producing AI, and the practices of the companies and governments that use AI." Finally, if you are an individual who uses AI, please do not interpret this as an attack against you: I do not know your circumstances or reasons for using AI, and therefore, I cannot fairly judge your use (or non-use) of it. This post is intended to be informational, not judgmental. Having said that, let's get started! Energy Use Asking ChatGPT a question uses approximately 10x as much energy as a traditional search engine1 (this excludes AI-supplemented search engines, like Google's "AI Overview" which incurs the energy use of a traditional search query plus the energy use of AI summarization). Indeed, AI uses so much electricity that the U.S. government wants to build more nuclear reactors.2 As it stands currently, AI represents a significant increase in energy usage,3 and most of that energy is extracted from fossil fuels.4 We know that burning fossil fuels is detrimental to the environment, and considering that we all live in the environment, we should be motivated to take care of it. In some instances, AI data centres are powered completely by renewable energy.

Unfortunately, unless the data centre builds new renewable energy sources to supply 100% of its power, this still results in an increase in fossil fuel usage. Why? Because if the data centre is using pre-existing sources for renewable energy, it is taking that energy away from other consumers that need it. To pick up the slack, we must generate more energy, and most of that generation is done using coal or natural gas.4 Additionally, many residents who live near AI data centres are seeing increased electricity bills due to the amount of electricity those data centres use.5 That's right, trillion dollar companies are using you to subsidize their power bill! Water Use AI uses a lot of water. This cannot be understated. It is estimated that AI alone will use between 4 and 7 billion cubic metres of water per year in 2027,6 and the amount of water AI uses is expected to increase. To put that number into perspective, on average someone in the U.S. uses 292 cubic metres of water per year. AI's water consumption will be equivalent to the consumption of 24 million people - that's 24 million people who didn't have to worry about getting fresh water before, who will have to worry about getting fresh water very soon. This water is needed for cooling the components used in data centres. These data centres are in direct competition with local residents (who also need fresh water), and, like with electricity, residents who live near data centres are seeing increases in their water bill. To make matters worse, most new data centres are built in areas that already have high water stress.7 Overall, these two issues (water & electricity) compound to increase cost of living in an economy where most people are already struggling, and we haven't even considered the knock-on effects of the water scarcity that this will induce. Data Centres Don't Actually Create (many) Jobs Many politicians and CEOs are claiming that AI development, such as building AI data centres, will be a boon to local economies are create lots of jobs. However, this is not strictly true: Building data centres results in a large boom in construction, but once the data centre is built, it typically employs a very small number of people.23 Communities are mislead into spending large amounts of money to build these data centres, and are essentially left holding the bag when the promise of "more jobs" falls through.

Mass Surveillance History has shown us that mass surveillance is bad. Unfortunately, numerous governments are actively using AI to enable mass surveillance on an unprecedented scale. Some notable organizations include:

The Federal Government of the United States, which recently attempted to strongarm Anthropic into allowing their AI to be used for mass surveillance.8 The Government of the United Kingdom, which has plans to use AI to "catch criminals before they strike."9 (sound familiar?) The Government of the People's Republic of China, which has the most advanced and ubiquitous AI-powered surveillance system.10 The Government of India, which uses AI for "predictive policing."11 The Federal Government of Russia, which uses AI to identify and track political dissidents.12

I'm sure there are more examples that I'm missing here (in fact, I would wager that every state on the planet is interested in the surveillance applications of AI), but this is everything I could find with good sources. While government surveillance is nothing new, what is new is the capabilities AI brings for capturing and analyzing surveillance data. What used to take hundreds of person-hours to analyze can now be analyzed much faster, with much less humans involved. What used to be reserved for "known dissidents" or "persons of interest" can now be used on everyone, just in case. Military Use As with most new technologies, the military industrial complex is interested in the applications of AI. The U.S. government openly wants to transition to a largely "AI first" approach for the Department of Defense, claiming that this transition "is a race".14 This has one glaring issue: AI has not yet been proven as a reliable source for data analysis (which is what the DoD primarily wants to use it for). Now, this definitely falls into the realm of opinion, but I believe that any technology that will influence decisions about whether or not people will live or die should be proven to work reliably. In other words, we should not be racing into adopting AI for military use. At best, this is irresponsible, and at worst, it could conceivably result in the loss of billions of innocent lives. (see: Nuclear Armageddon) Even if this doesn't result in the actual end of the world, (which admittedly is very unlikely) it is very likely that this will result in innocent people dying, who otherwise wouldn't have.

Adult Content Similar to the military, the adult industry is also very interested in how they can adopt AI. AI-generated pornography is becoming more and more common, and this is presenting many ethical and moral issues. AI is being used to make porn more addictive,15 create non-consensual explicit deepfakes,15 and create CSAM.16 These things are all bad. Copyright Infringement AI must be trained on large amounts of data. It is estimated that ChatGPT 4 was trained on over 1 petabyte of data (for reference, that's approximately 60 million selfies) and Anthropic destroyed millions of real books to train Claude.31 Unfortunately, most of this data is the intellectual property of other people. In fact, every large AI model today is trained from significant amounts of copyrighted data.17 AI companies like OpenAI or Anthropic, who are valued at billions of dollars, are profiting from the work of other people, and those other people are receiving nothing in compensation. AI Psychosis "AI Psychosis" is when a person gains new psychotic symptoms from AI use. While this is not a clinical diagnosis and is largely unresearched, it is a real problem that worsens pre-existing mental illnesses and can even induce delusions in otherwise healthy individuals.24 AI has been attributed to the death of many people, primarily by suicide. (I won't go super into detail here, but Appendix 1 of the cited source has several cases of AI psychosis) While the cause of AI psychosis is not known, it is speculated that it is related to the sycophantic nature of LLMs,2430 that is, their tendency to confirm the user's beliefs under any circumstance. A Society Where Nobody Works Many notable figures have spoken about how AI could enable a post-work society.18 A "post-work society" is one where labour is entirely optional, or even a luxury: no human being will have to work, and can dedicate their lives pursuing what they want most, because AI will take care of all the necessities. Sounds good, right? However, nobody - no person, no government, no organization - has agreed on how society will work once labour is no longer required. Currently, in most parts of the world, you need money in order to live, and you must work in order to earn money.

So if AI started doing all of the work... AI would have all the money, and the humans would all be poor! Clearly, this isn't tenable. So what do we do instead? Many things have been suggested, such as UBI, communism, or techno-feudalism, but as I said before, nobody has agreed on what we should do. No government has a plan for how they will transition to a post-work society. If we have powerful organizations trying in earnest to create a post-work society, we should have a plan for how the post-work society will function first, before we actually create one. A Society Where Nobody Learns AI is already negatively impacting how children learn, causing issues with cognitive, social, and emotional development.19 This is already bad. What's worse, however, is that if we don't curb AI usage by students in schools, we will end up with a generation of adults who are dependent on AI for most of their decision making and problem solving. And this isn't speculation - compared to people who don't use AI, people who frequently use AI underperform in cognitive tasks.20 A Surprise Guest Segment??? I asked one of my friends (who is an elementary school teacher) to share their thoughts on AI and education. Instead of sharing their thoughts, they sent me an entire essay! This next section is written by them.

I think my biggest thing and I know this is obvious but kids ability to think for themselves is super important. AI has become the easy way out for lots of students and they aren’t building resilience to work through challenges in their schoolwork, or in their daily lives for that matter. The second they struggle a tiny bit they turn to AI to do the work for them because they’re seeing the path of least resistance, which is incredibly human but also incredibly dangerous. If one day those tools do completely disappear, they’ll need to relearn (or learn from scratch) how to use their own brain and how to push through resistance that’s actually really important when learning. Humans need to have what we call a “productive struggle” in order to learn new things, and again that builds resilience. Which helps long term, not just with academics but with the ups and downs of life in general. So completely removing that struggle to rely on a robot that isn’t perfect, but treating it as if it is a perfect replacement for a human brain, is awful.