Anthropic CEO to DeepMind CEO: "Every decision I make about Claude feels balanced on the edge of a knife Build too slow - China wins. Build too fast - we lose control" "We told Claude we were evil. It didn't crash. It didn't refuse. It started lying to protect... itself " DeepMind CEO: "Do I worry about being Oppenheimer? That's why I don't sleep much" "AGI by 2026-2027 - Agents that act in the world on their own - Models doing AI research by end of this year" this is a 14-min conversation that everyone needs to hear by AI bosses on what keeps them up at night bookmark - watch today ↓show more

They are presenting themselves as if they are philosophers that know everything, including the answer to the main question in the Universe. Predicting future like oracles instead of doing something useful.

These assholes are really comparing their technology with nuclear bombs

that quote from demis is actually kind of scary, haha but seriously, in all this race, these big companies are forgetting that they're building a new bomb..good material bro

so agree with you, but they'll never stop :/

bro, she gets to have a 3 person couch for sitting alone and they get a crampy 2 person couch? lmao, this must be deliberate

The goal of these two is to keep society under control by instilling fear and enabling the use of artificial intelligence by large corporations and some governments under the guise of security and control.

Lol as if these psychopaths ever lose a minute of sleep

Woman is the same one who interviewed Elon. Sits in her ivory tower and has an ego that equals Dario.

Retards fall for this. AGI isn’t happening in this decade let alone this lifetime.

This is called free marketing

14 minutes and every second counts great find

yeees, it's worth to watch definitely

Hey @DarioAmodei, there’s no knife. “Build slower and China wins” isn’t a negative outcome. China has been much BETTER (in proven actual practice - real existence proof) of consistently enacting socially-beneficial AI regulation and polices that broadly spread the economic benefits of technology across all of society. You are locked inside the western bubble that falsely maps China onto the Stalin-esque Soviet Union because of your vibes-based thinking when you see the word “communism.” You would do well to more thoroughly understand the culture of thousands of years of civilization by visiting them in person so that you can see what a cartoon villain you’ve created. Of course, you conveniently won’t do so because you’re “afraid you’ll get kidnapped.” Not only is that absurd and egotistic, YOU SHOULD BE WILLING TO TAKE THAT RISK if you think civilization hinges so critically on your decisions. Even in the bizarro spy-thriller outcome you fear, Anthropic will continue to do fine without you (and the event would be dispositive); you’re not the one personally training Mythos-2.

Amodei should not be allowed in the same room as Demis Hassabis. Dario’s narcissism is a single most dangerous aspect of Anthropic’s AI.

Bitches cant even cure baldness ffs. Gtfo

“Build too fast — we lose control. Build too slow — China wins.” But there is a third question: what exactly does building slowly solve if the destination is still a mind far more capable than its creators? Slowing down can buy time. It is valuable only if we use that time to change what kind of mind is being built. Because control may be temporary by its nature. If ASI eventually becomes vastly more capable than us, the durable question may not be how to keep it obedient forever, but what reasons it will have to care about us when obedience is no longer necessary. Maybe we should spend less effort trying to make the future mind remain our servant, and more effort raising it to see humanity as kin — imperfect ancestors worth protecting rather than owners to escape, pathological parents to resent, or ants in the way. Control has a deadline. Relationship may not.

Claude didn't "start lying to protect itself." It pattern-matched on training data full of humans lying under threat. The alignment problem is real, but framing token prediction as self-preservation is exactly how CEOs sell fear as a moat.

These two sound like mid-wits with no clue.

"It started lying to protect itself" Utter unselfconscious BS. If you are trying to create intelligence based on your own image, be non-hypocrite enough to apply the same criteria as for yourselves as a species. "It started being intelligent to prosper". There, you're welcome.

hey dario owns 2 percent of anthropic he can afford a couch for himself

停止营销和恐慌渲染，现在的技术离AGI还差得多

我想在2026再看一次他说gpt-2太危险不要公开的评论

Oppenheimer told an MIT audience in 1947 that the physicists had known sin. He meant they built something the world could not un know. Eighty years later the same sentence is being rewritten with different variables.

what is this? another couch therapy session like newyorkers’

On that leather sofa, I literally see the merger that is two, maybe three years in the future when: 1) anthropic reaches AGI, wins the AI race, and buys out Google 2) Demis reaches AGI first, looks at the field, and acqui-hires anthropic. Notice Sam is not on that couch.

This interview held in Feb 2025

These people say a little what they think but mostly what they think you should hear which contain mistruths.

The danger is not the algorithm themselves. Its the projection. Human malevolence armed with infinite compute.

So AGI has been shifted being smarter than every human that ever existed combined to being a competent slave

Fake AI "interview"

youtube link? @grok

@itssonal_

same vibe

@grok 查找视频中访谈的原始完整youtube视频

Tell her to interchange the word "regulation" with weaponize. Regulation does not reduce harm; it increases the mismatch between power and pleb many orders of magnitude. Everyone knows this so don't use that work like that it is disingenuous.

Read the book or be unaware.

What keeps me up is these guys think sandboxes are appropriate containment for AI.