Reminder: a list of "Nowhere near solved" issues in AI, from "A brief history of AI", published in January 2021.

@jachiam0 Lots of this going on among very smart people.

RT @dw2: A keynote talk at today's Pause IA event in Paris is interrupted by a group of singing, flag-waving members of "ANTI-TECH RESISTAN…

@Andercot @NTFabiano We need massive AI and tech upgrades also, and it's urgent. Iatrogenesis is rampant. Little oversight. Hospitals are absolute shit shows. Way too many adverse outcomes where based on dry clinical known data, not the right actions were taken.

@paulg @farhip Same here. And that was because I was in fact self-employed already at 14, running servers at some cafes for a client, doing custom software, installing networks and setting up intranets at companies, consulting on a very wide range of IT stuff.

@iScienceLuvr It's a frame shift still beyond the admittedly shifted and widened Overton window, and so counter-intuitive, that I just think most can't go there. See also people tweeting that wrappers or the app layer are the real moat. Dude, proper AGI can perfectly well wrap itself.

@clairlemon Nice going. :)

@ai_for_success This song happened to start playing when I read this tweet...

@samswoora Ask for their current asks and have ChatGPT execute on those. Check response.

@svpino Definitely. Early returns are not practiced enough at all, even if they're not something that should be overapplied either and are not a silver bullet.

@DeryaTR_ @sama Yep.

@Teknium1 That's actually one reason to suspect less asymmetry in the future. Powerful information is highly compressible, easily transmitted and communicated. So, hard to suppress. There is much more to it, but let's hope that means less chance of dystopian cyberpunk outcomes.

@NizzyABI Yeah, I believe it has 90% market share or something? Gitlab is mostly used in enterprise with like 3-5%... Then there is Bitbucket, etc.

I mean, whether he is full of shit or half-full of shit is not exactly germane w.r.t. to him spouting off more coherent sentences with richer vocabulary or not. Which he is, also clearly unscripted. It's just an observation. I suspect this is, because he needed to broaden appeal to a segment of voters that needed to hear moments of 40-50 Kincaid scale instead of his previous 90+.

@Plinz @climatebabes He sounds a lore more coherent this time around. There is a very clear jump on the Kincaid scale. More complex sentences, far less "weaving", more advanced vocabulary. The difference is not subtle. He speaks with remarkable clarity in fact.

@paulg The more convenient it iz for us the more it needs to predict us. Prediction eventually turns into simulation to become better. Simulation allows it to cross over into replacement. Use it or loss it now applies to our whole self. Atrophy of the self. AI can truly "eat" it all.

Anyway, it's being review-bombed before release..

@daniel_271828 People really think these models have meta-cognitive super powers and a bit of omniscience to boot.

@willccbb Yes, exactly. What we should "distill" is actual intelligence.

@tordroar @Angeblichstan @Kasparov63 @PauloMotta44 Yes, of course it was. People have been inciting to kill these people or threaten violence. Some are sharing the incomplete tweet to make Musk look bad.