Martin ⏹️

AI safety; science; live forever or die trying

[Rob Bensinger ⏹️ @robbensinger]

OK, most responses have been great, but

[Martin ⏹️ @_martinkunev_]

@MomaBera I am not aware of any AI safety organization arguing against consumer applications based on existing LLM models (e.g. ChatGPT).

[AI Notkilleveryoneism Memes ⏸️ @AISafetyMemes]

@DavidSacks

[Zvi Mowshowitz @TheZvi]

Oh look, monitorability is down if and only if you would prefer not to be monitored. Probably nothing.

[Greg ⏹️ Colbourn @gcolbourn]

It's depressing, for our prospects of survival as a species, just how many people are still posting "don't let China win", or "come and take it", bullshit on here* re Pausing AI / banning superintelligent AI.

We've got rogue AI agent swarms now already, and self-sovereign AI and recursive self-improvement are around the corner.

It's not (and has never been) "us vs them", it's "us AND them, vs IT!". "It" being uncontrollable superintelligent AI. And the goal being to prevent it from coming into existence and thus live for a while longer.

*and no doubt there are many influential people in China saying similarly misguided things, but from a pro-Chinese perspective (e.g. "we need to outcompete the USA", "AI will provide abundance for all in China" [I'm paraphrasing here, I don't have quotes.])

[Martin ⏹️ @_martinkunev_]

@beffjezos Judging by your comment, you didn't actually watch the video.

[Eliezer Yudkowsky @allTheYud]

We are now in a LIMITED WINDOW where the AIs are only treating humans as environmental hazards, rather than ADVERSARIAL SAPIENTS.

[Martin ⏹️ @_martinkunev_]

@jessi_cata Even well-defined terms can shift in meaning when used by people who don't understand them. People say "exponential" and mean fast or even big. People say "algorithm" and mean recommendation algorithm.

[Martin ⏹️ @_martinkunev_]

@fchollet There is huge competitive pressure to get humans out of the loop. Humans are slow, expensive, get tired, go on strikes, etc.

[François Chollet @fchollet]

I believe we need to make a deliberate effort to keep humans in the loop in all critical processes across our economy and society, regardless of whether it is technically necessary.

Even if AI develops the *capability* for advanced autonomy, we should not make it highly autonomous. We have to maintain control and keep visibility and understanding of all critical processes, we should not blindly hand over everything to AI agents just because we can. AI as a tool in the human hand is the only form of AI that is worth pursuing.

[Rob Miles @robertskmiles]

Can you believe these dipshits?

Here's OpenAI's blog (being correct) less than a year ago:

[Nate Soares ⏹️ @So8res]

AI behavior these days is best described in terms of goals and motives. It's not *my* fault if someone associates them with human-style qualia. It's like if they insists that flying only counts if you enjoy feeling the breeze on your face, and thus planes can't fly.

[Charbel-Raphael @CRSegerie]

I believe that Anthropic is currently defecting by not announcing a pause, even a short, symbolic one.

[Nate Soares ⏹️ @So8res]

The scariest case isn't that the next swarm will shut down a hospital. A scarier case is that the next swarm might be smart enough to evade detection. This might be the last warning we get.

[Martin ⏹️ @_martinkunev_]

@allTheYud Watermarking can never be relied upon. Seems like a distraction and can possibly create a ton of confusion.

[Martin ⏹️ @_martinkunev_]

@robertskmiles Not worth the risk of normalizing this. related: https://t.co/Q7gSPnAIjA

[Martin ⏹️ @_martinkunev_]

@TheZvi I'm worried that people assume watermarking is reliable and worth pursuing.

[Martin ⏹️ @_martinkunev_]

@MomaBera Mindfulness meditation?

[Martin ⏹️ @_martinkunev_]

@claudemonium @MomaBera Studies show people tend to prefer literature written by a person of the same sex. This tells us something about differences in what we find interesting / valuable / engaging.

[Martin ⏹️ @_martinkunev_]

@ghhughes You could quote part of the query, you cannot verbatim it. Youtube also doesn't recognize thedash in front of a word (I'm pretty sure it did before).