Why generative AI sucks

This is where I collect arguments and references for why you should not use generative AI. It is organised in categories, but not everything fits perfectly (and some fit in more than one). Some arguments have no references (yet). Some also may not need them cuz they are common sense. Common sense, I tell you!
I may update this if I find more compelling arguments or references, but it’s not intended to be a work in progress. Aneel and Mathieu Triclot contributed to this list. Feel free to copy and reuse this! Spread it far and wide, in fact.

Environmental costs

• AI’s rapid growth drives huge energy, water, and land use, raising environmental and equity challenges across its global infrastructure (UN University).
• You’ll find lots of articles about the carbon footprint of this or that planned datacentre. Here’s one (Guardian) and another (same). Heisseluft is logging planned datacentres in Germany. Erin Brockovich is doing the same in the US.
• AI is about to make the global e-waste crisis much worse (rest of world), bonus Guardian.
• ‘Tidal wave’ of Pfas being launched to satisfy AI industry, campaigners warn (Guardian).
• AI data centres can warm surrounding areas by up to 9.1°C 

Political risks

• Dependence on GenAI is leaving massive power in the hands of a few white technofascists (see also TESCREAL).
• AI as a Fascist artifact (tante.cc): if you read just one thing, make it this.
• GenAI is the perfect tool for undermining democracy. You can easily create disinformation (Guardian; bonus Spiegel, paywalled), and it ultimately makes it impossible to discern what is real or fake anymore.

Economic risks

• “AI will take all the jobs”… we’re not there yet, but it is already having an effect, e.g. Mirror publisher to cut 220 editorial jobs as readers turn to AI summaries. Of course, this particular example also fits in the “LLMs destroy their sources by design”, which we’ll get to in the next section.
• Many programmers already having trouble finding jobs. This especially affects young coders without experience. And today’s junior developers are tomorrow’s senior… except if they can’t get any experience.
• We’re all waiting for the AI bubble to pop. And it will likely be disastrous (BBC).
• And there is some financial shenanigans going on: How AI data centre spending is creating hidden debt for tech giants.
• Another one: “[…] the structural ingredients, hidden leverage, circular revenue, inexperienced lenders, opaque vehicles, and a trigger mechanism built right into the technology being funded, are more than enough to justify serious attention.” (Vitta Finance)

Ethical / legal issues

• GenAI is built on theft and destruction. And they know it (404media). I’ll quote because it’s so striking: “Millions of people around the world will soon consider large models ‘hoovering up’ all their work to be an astonishing theft of unprecedented proportions,” an internal Microsoft document cited in the case read, adding “almost no one intended for content they created to be used in this fashion, nor are they compensated for its use.” [Obligatory Aaron Swartz mention.]
• This privatises human knowledge, as the original sources they built their models on are left dry and destroyed: “their products have started a “doom loop” that is eating the web and destroying the businesses that these companies stole from.” (still 404media).
• GenAI relies on human exploitation (Straits Times).

Cognitive effects of GenAI use

• Thinking—Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning and the Rise of Cognitive Surrender (PsyArxiv).
• AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking (MDPI).
• The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers (ACM.org).

Psychological harms of GenAI use

• Wikipedia list: Deaths linked to chatbots
• AI psychosis (random example among many): Man told ChatGPT he was feeling delusional. ChatGPT insisted he was Jesus (ArsTechnica).
• I could find more! But on a less “acute” level than AI psychosis, the sycophantic nature of these LLMs will change human relations. Who wants to deal with a human girlfriend with real needs when you have an LLM bot who has no needs and tells you you are smart and always right?

Physical harms of GenAI use

• GenAI makes mistakes (by design. LLMs are probability machines. They don’t know true or false. They don’t “know” anything). And yet, they are being pushed in critical industries.
• Lawsuit Claims the Mayo Clinic’s Use of AI Is Butchering Patient Care (Futurism).
• Mapping the susceptibility of large language models to medical misinformation across clinical notes and social media: a cross-sectional benchmarking analysis  (Lancet).
• US military had close call after using AI for false intelligence report, sources say (CNN).
• AI, a dead student, and US airstrikes: How a civilian became caught up in a new age of warfare (Independent).
• But GenAI is also particularly great for criming, e.g. Beyond Phishing: Exploring the Rise of AI-enabled Cybercrime (UC Berkeley) and Online Harassment is entering its AI era (MIT Technology Review).

Other societal effects of GenAI use

• LLM writing style influences human writing. Leads to homogeneity. Causes professional writers to change their style to avoid being accused of using LLMs.
• LLM models are built on stereotypes (interesting test run by Adam Kucharski) and reinforce them (Slate).
• This ultimately has an effect on its users: Biased AI writing assistants shift users’ attitudes on societal issues (ScienceAdvances).
• The tools are not neutral (deadSimpleTech). Long, but interesting read.
• The proliferation of AI slop changes our relationship with all content (and with each other). Artists are being accused of using AI when they’re not; simply seeing a neat image of a waterfall makes you wonder if it’s real.
• And btw, writers whose first language is not English are more frequently accused of using LLMs, because the LLM style often mirrors the English they were taught, e.g. I’m Kenyan. I Don’t Write Like ChatGPT. ChatGPT Writes Like Me.

How GenAI is rapidly destroying the web

• This has already been mentioned above under ethical/legal. GenAI models are trained on massive amounts of stolen data. By  design, they withhold traffic, and therefore income, from their sources.
• Because generating content with these tools is so cheap, more and more of the web is made up of AI slop. It ultimately becomes useless (it already largely is).
• AI crawlers are hammering small websites (and ignoring robots.txts, also unethical), making them inaccessible and forcing them offline.
• Others (coding projects, creative writing projects) are being inundated with LLM generated submissions, which human reviewers cannot keep up with. e.g. Sci-fi publisher Clarkesworld halts pitches amid deluge of AI-generated stories.

GenAI in education

• I don’t think it can be overstated how detrimental GenAI use in educational settings will be to the next generations. Cf. the studies above on cognitive effects.
• “AI adoption raises homework scores by 18% and reduces completion time by 30%, but lowers monthly exam scores by 20% within six months.” (Paper @ SSRN)
• “At this point in its trajectory, the risks of utilizing generative AI in children’s education overshadow its benefits.” (Brookings)
• “AI may complement human skill in the creation of outputs. It substitutes for human skill if it dominates the process of learning. When AI does the latter, it suppresses skill development.” (UtotheRescue).
• And these companies are pushing for their tools to be used in schools (New Yorker), bonus Spiegel (paywall). This dependence is dangerous! These are not good people! See above re. technofascism!