A Lot Has Happened in A.I. Let’s Catch Up.

Рет қаралды 16,465

5 ай бұрын

Thursday marked the one-year anniversary of the release of ChatGPT. A lot has happened since. OpenAI, the makers of ChatGPT, recently dominated headlines again after the nonprofit board of directors fired C.E.O. Sam Altman, only for him to return several days later.
But that drama isn’t actually the most important thing going on in the A.I. world, which hasn’t slowed down over the past year, even as people are still discovering ChatGPT for the first time and reckoning with all of its implications.
Tech journalists Kevin Roose and Casey Newton are hosts of the weekly podcast “Hard Fork (www.nytimes.com/column/hard-fork) .” Roose is my colleague at The Times, where he writes a tech column called “The Shift. (www.nytimes.com/column/the-shift) ” Newton is the founder and editor of Platformer (www.platformer.news/) , a newsletter about the intersection of technology and democracy. They’ve been closely tracking developments in the field since well before ChatGPT launched. I invited them on the show to catch up on the state of A.I.
We discuss: who is - and isn’t - integrating ChatGPT into their daily lives, the ripe market for A.I. social companions, why so many companies are hesitant to dive in, progress in the field of A.I. “interpretability” research, and America’s “fecklessness” that cedes major A.I. benefits to the private sector, and much more.
Recommendations:
Electrifying America (mitpress.mit.edu/978026264030...) by David E. Nye
Your Face Belongs to Us (www.penguinrandomhouse.com/bo...) by Kashmir Hill
“Intro to Large Language Models ( • [1hr Talk] Intro to La... ) ” by Andrej Karpathy (video)
Import AI (jack-clark.net/) by Jack Clark.
AI Snake Oil (www.aisnakeoil.com/) by Arvind Narayanan and Sayash Kapoor
Pragmatic Engineer (blog.pragmaticengineer.com/au...) by Gergely Orosz
Thoughts? Guest suggestions? Email us at ezrakleinshow@nytimes.com.
You can find transcripts (posted midday) and more episodes of “The Ezra Klein Show” at nytimes.com/ezra-klein-podcast (www.nytimes.com/column/ezra-k...) , and you can find Ezra on Twitter @ezraklein. Book recommendations from all our guests are listed at www.nytimes.com/article/ezra-... (www.nytimes.com/article/ezra-...) .
This episode of “The Ezra Klein Show” was produced by Rollin Hu. Fact checking by Michelle Harris, with Kate Sinclair and Mary Marge Locker. Our senior engineer is Jeff Geld. Our senior editor is Claire Gordon. The show’s production team also includes Emefa Agawu and Kristin Lin. Original music by Isaac Jones. Audience strategy by Kristina Samulewski and Shannon Busta. The executive producer of New York Times Opinion Audio is Annie-Rose Strasser. And special thanks to Sonia Herrero.

Пікірлер: 40

@buzzwashere1 5 ай бұрын

Great episode! Please revisit this topic regularly.

@dougirvin2413 5 ай бұрын

Great show Ezra! I'm a retired Corrections Officer. Towards the end of my career my Department adopted a new prisoner count program ( very important software for a prison!) It was terrible! One day I was complaining about it when my Captain walked into my cell block and pointed out that all the computer science grads who got good grades either went to work in Silicon Valley or Wall Street so who did that leave us with?...Good point Capt.!!!

@TheWingjammer 5 ай бұрын

happy you are back Ezra!

@chrisrecord5625 5 ай бұрын

Well, there is sure to be a George Santos bot that is your personal assistant, therapist, and comic relief even if he is in jail.😉😉

@speciesofspaces 5 ай бұрын

Odd conversation in the sense of "I keep waiting for it to..." but the fact is the changes are happening and not only that but are being deployed. I worry some of us are looking at this through the Hollywood narrative lens and not the more mundane constantly experimenting and into the wild lens where being "convinced" isn't from any one thing or even a single category of a big thing but instead more like automobile advancement where the subtle changes have profound societal impacts and unintended consequences. In this case it is much, much faster and less about a mechanical difference of the future and one about cognizers and what that means for humans and machines etc. This last point is still not so well established which is cognizers as they do come in all shapes and sizes from simple living examples to more complex ones is a bridge between where we normally think the living and nonliving share space and communicate from. The "assistant" is first gen for sure but from the looks of it in a few short years the assistant could very well be the associate etc. But as was remarked the "embodied intelligences" folks are still around and it is also still emerging in the same AI field etc.

@videob1962 5 ай бұрын

I think you're wrong about China's (lack) of AI progress. Baidu's model is multimodal, and outscores GPT-4 in some benchmarks. Its just not reported in the west, and some reports about it are supressed and / or censored.

@PaulHigginbothamSr 4 ай бұрын

This is exactly what one would think would happen as ai reaches near human capability. This is a surface kerfluffle. Meaning not critical to ai itself. When llms reach beyond human capabilities in some aspects but not others means it does not yet have a human personality. By that I mean what the level of the unit is capable of. Very limited in scope really but when all the internet is fitted into rapid access ability many things happen to make the human think it has feelings it is incapable of, like cat owners anthropomorphising it's love and thinking ability. Remember a cat will consume you if food is unavailable.

@EricJohnson-lo7mj 5 ай бұрын

Expect an AI-enabled black swan event to happen sometime, if not sooner then later. LLMs exhibit far too many edge cases to make them reliably "safe". This problem will amplify as the number of parameters increases; in the near future, the number of parameters in LLMs will increase exponentially. Under these conditions, the prospect that an unsafe output will escape the safeguards in such systems is a mathematical certainty. Many ways to deliberately circumvent such safeguards have already been demonstrated. The real existential risk does not necessarily come from LLMs themselves, but in how such systems can amplify the misapplication or misuse of other technologies. The rule of unintended consequences plays a large part here, not just diabolical pots by supervillians. Combine ever-larger LLMs with other AI models such as AlphaFold... I think that's enough AI nightmare fuel for now.