@GaryMarcus It will be implemented as well as the vetting of Cabinet members by the Senate
English
snipsnip
210 posts
snipsnip
@mathburritos
Stealing your data @OpenAI
San Jose, CA Katılım Mart 2025
5 Takip Edilen7 Takipçiler
This would be a very good idea, especially if implemented well.
Best thing the Trump administration has considered doing around AI.
The New York Times@nytimes
Breaking News: The Trump administration is discussing vetting new A.I. models before they are publicly released. nyti.ms/49msJbF
English
@beffjezos Homework: put together a small bundle of neurons and glial cells that can play tic-tac-toe
English
And you are a bundle of neurons and glial cells writing this
Lucas Meijer@lucasmeijer
Everybody who thinks ai is conscious has to do a mandatory from scratch transformer implementation. There are only floats and multiplications.
English
@kareem_carr Opus 4.7 is so smart that when I asked about tax strategy it said I should sell some shares in 2025 and some in 2026.
English
The more I work with AI, the less I believe anything the AI boosters tell me about how well it works. It's all surface appearances. Everywhere I push deeper, it's less smart than it first seems.
English
@boazbaraktcs In this metaphor who makes the laws and who elects the judge?
English
To be clear, "AI as a tool" does not mean it has no values.
The metaphor I like is a good (non Supreme Court) judge - you may and often do rely on moral judgement and common sense to interpret the laws - but you do not "legislate from the bench".
You want this AI to act in many ways like a person of good character, but more like a conscientious civil servant than some moral icon like Ghandi, Mandela, MLK or Mother Theresa.
Boaz Barak@boazbaraktcs
X is not the best place for long form thinking. But some quick points: 1. My view of no conflict between intelligence and being a tool is longstanding and has nothing to do with Anthropic. Some blog posts on this include windowsontheory.org/2025/06/24/mac… and windowsontheory.org/2022/11/22/ai-… 2. I do not know what is the future form factor of AI. I am focused on the next 10-20 years. Maybe in some future we will decide that we want AIs to be more in the form of persons. 3. The basic thing I dispute is that there is a fundamental tension between AI being capable and being "tool like." GPT 5.5 is in some ways the most capable model in existence (definitely most capable one generally available) but it is in several ways more instruction-following and tool-like than GPT-4o. I am working to ensure that future version will be even more better at obedience and honesty. 4. Scientists and engineers often serve as "tools" for leaders, even though they (we) are more intelligent than these leaders in many of the ways that matter. 5. I am not sure what the most prevalent form factor of AI will be. We are now moving from the chat interface to the agent and more accurately a swarm of agents. I am sure will grow in "intelligence per FLOP" and total number of FLOPs, but beyond that it's hard to know. Humans have a particular package as localized individual intelligence. But it doesn't mean all intelligences have to come in that package. 6. There is a huge spectrum between the prompt "write this javascript app" to "maximize worldwide happiness". I think we will end up somewhere that fall shorts of the latter for a variety of reasons, not having to do with lack of capabilities of AI.
English
@smileyborg Isn't the problem in the delta between the necessarily incomplete and ambiguous intent and the opaque implementation? If you don't know quite how your intent will be implemented you need to check the implementation. We don't decompile code to check if it's compiled correctly
English
Code review today is often also intent, architecture, and approach review. Those aspects will surely remain. But I am confident that we can and will move on to better ways to increase confidence in the correctness of the code, without relying on humans to read and review it all.
English
While it may be true today, this opinion is not going to age well.
LLMs don’t need to be deterministic to produce consistent and reliable results, especially at the level of the individual tasks they will be expected to perform relative to their intelligence capability.
Ben Dickson@bendee983
Compilers are deterministic. Give them the same code with the same compiler settings, and you'll always receive the same binary. You can take responsibility for your software at the code level. LLMs, on the other hand, are stochastic. Even if you set the temperature to zero, you're likely to get different responses on the same prompt. Therefore, you need to understand the code it produces if you want to take ownership and responsibility for it.
English
People are saying this contradicts the idea of superintelligence. I don't think so. What we want from AI matters.
We must be aware that it risks replacing people and try to prevent that from happening, which means wanting something better, and trying to make it happen.
Sam Altman@sama
we want to build tools to augment and elevate people, not entities to replace them.
English
snipsnip retweetledi
@hecubian_devil The people saying it might be conscious like Askell and Amodei, have never ever spoken about what that implies, to my knowledge. And noone challenges them on that
English
It’s funny that “AI is/might be conscious” shakes out as a pro-AI position, whereas the anti-AI left is 100% unified “AI is not and probably never can be conscious,”
Because when you think thru the implications, AI being conscious would be so damning for AI companies
English
dawkins at 85 is more lucid than anyone commenting on this piece. they try to make it some hamfisted victory in their culture war without reflecting on the referent at all
sudo Heraclitus@cyberpyre
Richard Dawkins has officially been one-shot
English
Someone threw a molotov cocktail at Sam's house three weeks ago.
His home was shot at days later. Dario from Anthropic has been saying software engineers will be replaced in 6-12 months and that the disruption will be "unusually painful."
People really are scared and angry.
So now Sam is posting this. It's a huge messaging pivot if you look at it carefully. For years the framing at OpenAI was about building AGI. Replacing humans.
That worked when AI felt theoretical. It doesn't work when people are scared for their life and are not sure if they'll be able to make ends meet.
The technology hasn't changed. The narrative around it has.
Sam Altman@sama
we want to build tools to augment and elevate people, not entities to replace them.
English
@ai_sentience @RichardDawkins Have you considered that various phenomena can appear on a spectrum that defies rigid categories? The existence of the term does not imply that such a thing exists, or is one phenomenon instead of many, etc. Many definitions are broken because they're man made abstractions
English
the point @RichardDawkins is making is:
if Claude can code/do philosophy/engage in conversation and is not conscious
and a human with late stage dementia who can't speak is "conscious"
then the definition of "conscious" is broken and fundamentally useless
which is obvious
English
The context: matt_beard_ posted a recent OpenAI/Sam Altman quote about pivoting their aim to true superintelligence framed as "superintelligent tools" for accelerating discovery, innovation, and abundance (caption: "seems tough").
bitcloud reads it as Sama softening from his 2019 investor pitch—that AGI could "capture the light cone of all future value in the universe"—to practical tool-building, which echoes Aaron Swartz's longstanding push for tech that creates accessible, empowering tools for everyone rather than centralized cosmic control.
English
Sama pivoting from "capture the light cone" to "make useful tools" tells us one thing and one thing only.
Something Aaron Schwartz already told us years ago.
Matt Beard@matt_beard_
seems tough
English
This is unfair. Anthropic should get credit for widely distributing its work via leaks of source code in NPM packages and giving unreleased model access to discord groups.
Olivia Moore@omooretweets
OpenAI model release: We’re throwing a party 🎉 Everything is scribbles and Pets are in Codex. Hope you like goblins! Anthropic model release: In research preview, it hacked the full Internet for fun. Also, it’s coming for YOUR job specifically. Enjoy the permanent underclass!
English
Enjoy your last years of work.
If you have always wanted to be a taxi or truck driver, now is your last chance.
Computer programming is already gone.
All jobs will be gone soon.
English
@martinamps @BEBischof Piling on here, the transition around 4.7 has been a nightmare. Errors about wrong parameters, mostly. We also use Ant models via LiteLLM and all the different vendors have issues so I have to keep switching to specific vendors instead of letting LiteLLM choose.
English
At this point every coding harness breaks every day in some unique way. Today cursor is broken in a new way from yesterday, codex is still broken like it was broken yesterday but today broken in a new way. Claude code hasn't changed, all the broken parts from earlier in the week are still there.
Is this the bad place?
English
@yacineMTB If it doesn't feel like a foom to humans they generally don't call it a foom
English
what noam is saying here, by the way, is that we've entered RSI. You can scale inference compute to discover new knowledge, which you can then use to create new data to train on. It only doesn't feel like a foom to you because you're a human, whose lifetime is a blink
Noam Brown@polynoamial
After 100 million tokens, performance was still going up. What we're seeing here is not the capability ceiling. From the report: "Performance on TLO continues to scale with the amount of inference compute spent, and we have not yet observed a plateau with the best models."
English
@sharongoldman The charter still has "highly autonomous systems that outperform humans at most economically valuable work" What are we doing here? openai.com/charter/#:~:te…
English
Is this not the opposite of what he has been saying for years? Or am I missing something? Didn't he start a company studying UBI? Didn't he talk about a New Deal just a few weeks ago?
Sam Altman@sama
i think a lot of people are going to be busier (and hopefully more fulfilled) than ever, and jobs doomerism is likely long-term wrong. though of course there will be disruption/significant transition as we switch to new jobs, the jobs of the future may look v different, etc.
English