AndyXAndersen

Computer vision engineer, math phd. Interested in AI, science, ethics, society topics.

[Johnny Midnight ⚡️ @its_The_Dr]

Oh this is so funny especially since it’s his wife Scarlett Johansson. 😂🤣.

[AndyXAndersen @AndyXAndersen]

@ai_sentience Inelligence is a very small fraction of what is sentience. Bacteria are sentient, but not intelligent.

[AndyXAndersen @AndyXAndersen]

@tomfgoodwin I think this is well-understood.

People are excited because coding assistants really help with productivity, and likely the advances will continue, and not just in coding.

[AndyXAndersen @AndyXAndersen]

@DavidSKrueger Nothing meaningful will happen. Regulation and safety are going to to be utility-level reforms, at some point. Doom arguments won't be credible as they never were.

[AndyXAndersen @AndyXAndersen]

Nowadays it is like: Claude: fetch and edit and push.

[zek @zekramu]

it blows my mind that people are still using git.

[AndyXAndersen @AndyXAndersen]

@benjamin_horne Nope. They will happen. OpenAI and Anthropic have very good fundamentals and provide real value.

Sadly, even SpaceX will make it, and likely become a disappointment sooner rather than later.

[AndyXAndersen @AndyXAndersen]

@davidmanheim OK, looks like I misinterpreted the whole: "Cannot agree to disagree". The theorem implies that the bots will end up converging onto something.

[AndyXAndersen @AndyXAndersen]

@davidmanheim Interesting. Any time Anthropic comes with some arcane deception experiment, doom people get all worried.

This inability to compromise appears to me the ultimate proof for faux-bayesian-minded folks that doom is nigh. No?

[AndyXAndersen @AndyXAndersen]

@Grady_Booch @claudeai Why keep on reminding Claude it is not an I?

Surely bots are not conscious. But Claude has reasonable self-awareness based on reading its own stuff and running tools, and has reasonable competence.

So a simulated person one can work with, even if not human or conscious.

[AndyXAndersen @AndyXAndersen]

@emollick Genuinely puzzle by the attitude. Beter delve first in the post?

[AndyXAndersen @AndyXAndersen]

@ai_sentience Nothing is ever obvious. You have to do the heavy lifting. Be intellectually serious.

You are not building a solid reputation by claiming consciousness so early, and you won't be taken seriously when you actually may start having a point.

[AndyXAndersen @AndyXAndersen]

@AlphaMFPEFM @VraserX Pure LLM was never the goal. Or pure scaling. That is what the excitement was about, as were the new shiny pieces that were delivering, but the goal was always a system. OpenAI and the rest did whatever worked.

[AndyXAndersen @AndyXAndersen]

Seeing Elon as the man one can't never bet against is a recipe for disaster. His true believers are due for a rude shock, as the dude moves further and further from the approaches that made him successful.

[Kirsten Grind @KirstenGrind]

NEW: Ahead of the SpaceX IPO, we tracked hundreds of promises that Musk has made over the years (FSD, Mars etc). His success rate is…not good. And it’s getting worse. Some actual data in our months long analysis: https://t.co/u3McNwUE80 

[AndyXAndersen @AndyXAndersen]

@davidmanheim Yes, talking about agents too, not LLM. The point I think seems to be that not being able to agree to disagree does not mean doom. Does not mean easy to find compromise either.

[AndyXAndersen @AndyXAndersen]

@VraserX The tech is certainly evolving in a way not aligned either with their own preferred directions or their expectations.

Funny however that after criticizing all as band aid, now they say the band aid is what they claimed was needed all along.

[AndyXAndersen @AndyXAndersen]

@davidmanheim If by tools you mean smarter agents, yeah. May need to give the agents extra abilities. Once they air their differences, they need to learn to go forward with an imperfect solutions.

[AndyXAndersen @AndyXAndersen]

@davidmanheim I actually got the point just fine. The question is, once actors are dead set on their opinion, can they move forward? That is the crux.

[AndyXAndersen @AndyXAndersen]

@Dr_Gingerballs This is actually a good thing. One must calibrate the strategy, or go bust. That's how Anthropic will do well. Be bold, but then adjust.

[AndyXAndersen @AndyXAndersen]

@PeterDiamandis The problem is that SpaceX's ambitions for datacenters in space and cities on moon are not going to have ROI.

SpaceX is also behind in AI. It also competes hard with Oracle and others on serving compute.

Elon does not have a serious business model.

[AndyXAndersen @AndyXAndersen]

@dwarkesh_sp @jacobkimmel Same as with chess. Youth is gold. You need a mean lean brain to make discoveries. Insight is important, but you have to endurance to go down deep all the side alleys.