source avatarJames Altucher

Share

11 AREAS WHERE AI IS WEAK AND ALWAYS WILL BE - and the opportunities that will come from that. I love AI. I use it almost all day every day. I am 100% convinced the $10 trillion being spent on the infrastructure will be paid back in quadruple over the next ten years (more on this in another post). I am a super-bull on AI. But it’s important to be honest about where AI IS WEAK. And why this will create enormous enormous business opportunities. 1) AI IS IN A RELENTLESS PURSUIT OF AVERAGE – If you ask an AI “1 + 1 = ?” the correct answer is the average answer … “2”. In most areas of life, average is the best approach. And over billions of pieces of data in any particular domain, the pursuit of average is desirable. But for specific tasks, (design, good writing, scientific discovery that is not bulk experimentation, some aspects of diagnosis, marketing, etc etc) AI is great but not great enough. 2) MATH. Despite my example above on 1+1, most problems are harder than that. Code is good. AI is average. Calculators > AI. 3) DEMENTIA. Sometimes, with age, the human brain just fills up. There’s too many memories and experiences. Humans are ok at filtering through billions of experience to remember what is important. But ultimately humans start to fail at this. AI fails a lot faster. Talk to the #1 AI model in the world on a task for more than two days and the words “I’m sorry. I slipped” will be a frequent response from GPT 6 Astra or Claude Fable 5.2. ALL developers know this. 4. TASTE. Taste can be learned. This is a great benefit of AI. But GOOD TASTE IS AN OUTLIER. There’s a billion essays on the Internet. But there’s only 1000 “well written" essays. I put in quotes because even that phrase is hard to define. AI can learn when there are a billion examples. But much harder when its a 1000, and even those 1000 are determined by subjective opinion, 5. PROACTIVE AMNESIA: (think Drew Barrymore in the movie “50 first dates” - she wakes up every day with the same past memories). An AI model is frozen in time from when the model finished its learning. Only web seach and additional code can keep it updated 6. CONFIDENTLY WRONG, COMFORTABLY NUMB: In 2023 a New York lawyer filed a brief citing six court cases that ChatGPT made up, complete with fake quotes. The AI presented them as calmly as real ones, and the judge sanctioned the lawyers. 7. INCONSISTENT. AI can change its mind. And this is a great, human thing. But not always the best solution. Give an AI the same insurance claim twice and it may approve it once and deny it once. Ask it to grade the same student essay on two days and you can get a B+ and an A−. Variety is fine for brainstorming. For decisions about money, grades or medical care, the same facts must always produce the same answer. 8. THE LONG TAIL. AI knows the rare cases exist (e.g. good writing), but it bets on the common answer. The costly mistakes live in unusual combinations that no dataset covers: a rare medical condition, for instance. Or exceptional design to make your brand stand out. 9. EAGER TO PLEASE. AI will mirror your hope. This is why so much “prompt engineering” starts off with, “Tell me I’m wrong.” Tell an AI “I think this mole is nothing” and it tends to reassure you. Tell it “I think my landlord can’t keep my deposit” and it tends to agree. Costly for doctors, lawyers, etc. 10. GRADES ITS OWN HOMEWORK. Ask an AI to write code, then ask it to check that code, and it usually says the code is fine, including the bug it just wrote. Btw, AI happens to be GREAT at code but only because this is what ALL of the AI companies have decided to focus on, with trillions of pieces of data. And yet it STILL slips for this reason. It just happens to be the source of 99% of AI revenues so they focus on this one domain. 11. GETTING WORSE IN EDGE DOMAINS. . AI labs are putting most of their effort into coding. That’s where the big revenues are. But it takes a 10 trillion parameter model to keep the edge. Its a race to the bottom. And every bit spent getting better at code is a bit not spent on other domains: the taste of a great essay, or the nuances of a niche industry. Niche domains are getting ignored, and new models sometimes get worse in them. Focusing on solutions, instead of purely method, is where the next advance has to come from. More on this later. But do not think that opportunity has left you behind. It's only just beginning.

No.0 picture
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.