AI Might Be Dumber Than a Cat, but It's Still Transformative
Does it even matter how smart AI is?
Greetings from Miami, where I'm attending Adobe Max to check out the company's new AI toolbox for creators. I'll have more to say about its Firefly Video Model, Generative Extend in Premiere, and all the other ways it's spreading AI through its creative-software empire in Thursday's newsletter after I've had a chance to absorb the entire show. But if you want a recap, The Verge has a pretty good one.
In any case, What's happening at Max is a pretty good indicator of where the world is at with AI: We're beyond the novelty and the experimentation, and now we're deep into the integration with real work. I find it strange, then, that so much of the oxygen in the room is being sucked out by influencers and players prognosticating about our AI future — be it light or dark — when there's so much actual work getting done.
More on that in a second, but I'd like to take a second to welcome readers from The Boston Globe, MSNBC, The Telegraph, The Daily Mail, CNN, and everybody I've met at Adobe Max. I appreciate you all, and look forward to hearing from you in the comments or on LinkedIn.
One more note: I also recently appeared on the AI for Everyone podcast with Myles Dhillon. This was recorded shortly after Ted Chaing's excellent New Yorker piece on whether or not AI will ever make art — aptly timed, considering the conversation below. It's a quick listen, and I highly recommend subscribing to Myles.
A message from Incogni
Keep Your SSN Off The Dark Web
Every day, data brokers profit from your sensitive info — phone number, DOB, SSN — selling it to the highest bidder. And who’s buying it? Best case: companies target you with ads. Worst case: scammers and identity thieves.
It's time you check out Incogni. It scrubs your personal data from the web, confronting the world’s data brokers on your behalf. And unlike other services, Incogni helps remove your sensitive information from all broker types, including those tricky People Search Sites.
Help protect yourself from identity theft, spam calls, and health insurers raising your rates. Plus, just for The Media Copilot readers: Get 55% off Incogni using code COPILOT.
Why AI Reasoning Debates Miss the Point
Given the news over the past week, it's totally understandable if you're suffering from AI whiplash.
Just a few days ago, AI pioneer Geoffrey Hinton won the Nobel prize in physics for his work on machine learning and artificial neural networks. Hinton has become something of an AI doomsayer in recent years, predicting that, as AI gets more powerful, it will likely lead to calamitous effects on the labor market and even democracy itself.
A few days later, however, one of Hinton's professional peers, Yann LeCun — Meta's chief AI scientist — gave an interview where he asserted that the current crop of large language models (LLMs), even newer "reasoning" models like OpenAI o1, are "dumber than a cat." The models only happen to be good at predicting text, LeCun said, and that they can't really reason or plan. LeCun believes that much of the concern about AI posing a danger to humanity, which Hinton frequently expresses, is largely exaggerated.
And yet here we have the CEO of Anthropic, Dario Amodei, waxing poetic about the transformative potential of AI across all kinds of human endeavors — from neuroscience to economics to the meaning of life — for 15,000 words. Amodei's vision might be naively optimistic, but it necessarily supports the view that today's LLMs are the foundation on which world-changing AI will be built upon.
Which brings us to Apple, and its just-released study, which concluded that LLMs don't actually reason at all. The study covers a lot of detail about the common things that trip up today's models, like adding irrelevant details to math problems, which even confuse supposed "reasoning" models like OpenAI's o1.
So things are confusing! AI might either be soon capable of surpassing our abilities in most, if not all, aspects of human endeavor, or a total paper tiger, never quite understanding well… anything, really, despite its ability to create poems about Tide pods whenever we want.
The truth is very little of this debate is relevant to what's actually happening in the working world with respect to AI. The back-and-forth over what is AGI, when superintelligence will emerge or whether it will emerge at all, and the semantics on reasoning are largely a distraction. They have nothing to do with what AI can do for a business or an individual right now, which is a lot.
Keep reading with a 7-day free trial
Subscribe to The Media Copilot to keep reading this post and get 7 days of free access to the full post archives.