Sign up to see the future, today<\/p>\n
Can\u2019t-miss innovations from the bleeding edge of science and tech<\/p>\n
Anthropic is doing what Anthropic does best: putting on its most concerned face so it can convince you they\u2019re the good guys in the industry.<\/p>\n
Last week, the world\u2019s most valuable AI startup revealed a new commercial<\/a> that took an unorthodox approach to promoting its technology. Rather than diving straight into bubbly optimism, it starts with a series of grim images as we hear voiceovers from different speakers ask some hard-hitting questions about AI.\u00a0<\/p>\n You couldn\u2019t blame someone for thinking they were watching a PSA. It starts with a burning house \u2014 which always bodes well. When someone asks, \u201cCan AI be trusted?\u201d we\u2019re shown what looks like surveillance footage with an AI algorithm scanning the faces of everyone in a crowd.<\/p>\n The most striking sequence comes right after.<\/p>\n \u201cWho\u2019s going to hit the brakes if we need to?\u201d another voice says. Then it cuts to an image of a cemetery lined with hundreds of headstones.<\/p>\n You don\u2019t need to ask Claude to explain this one to you. Anthropic is clearly acknowledging the fear that AI could kill us all.<\/p>\n But this is actually a feel-good story \u2014 at least in the eyes of Anthropic. The tagline, \u201cThere\u2019s hope in hard questions,\u201d assures the viewer that their concerns over AI \u2014 like its potential to destroy jobs<\/a> or erode or ability to think<\/a> \u2014 are all perfectly valid, but more importantly, that these are questions which Anthropic\u2019s leaders are keeping themselves up at night grappling with.<\/p>\n \u201cPeople have a lot of hard questions about AI,\u201d an announcement<\/a> from the company states. \u201cIt\u2019s our job to address them.\u201d<\/p>\n The commercial, which aired during the World Cup quarterfinal clash between Argentina and Switzerland over the weekend, quickly drew criticism online. One viral post mocked<\/a> the thinking behind the ad, writing, \u201cWhen we raise the question of stopping a dangerously powerful superintelligence we show 300 American gravestones for half a second.\u201d<\/p>\n Even OpenAI CEO Sam Altman fired a shot.<\/p>\n \u201ci thought this was satire,\u201d he wrote<\/a> in his typical all-lowercase style.<\/p>\n The ad is Anthropic in a nutshell. A key part of the company\u2019s mythology is that it was founded by former OpenAI employees who wanted to focus on safe AI development<\/a>. CEO Dario Amodei claims that he could\u2019ve been the one to release a chatbot that changed the world instead of OpenAI with ChatGPT, but held off because he was too concerned about the tech\u2019s risks.<\/p>\n Amodei has also been strikingly forward about AI\u2019s risks, saying it could wipe out half of all entry-level white-collar jobs<\/a>, while dangling the possibility that its models may already be conscious<\/a>. Recently, the company even called<\/a> for a global \u201cpause\u201d on AI development because it feared the technology could spiral out of human control. (It also hired an economist<\/a> who said that a 33 percent risk of human extinction from AI was acceptable.)<\/p>\n Those may seem like counterproductive things for the creator of one of the most widely used chatbots in the world to say. But it all serves a very deliberate point. By being the ones that doomsay about AI the most, Anthropic can also present itself as the only company that can be trusted to develop it<\/a>.\u00a0<\/p>\n Its actual track record suggests otherwise. In February, it dropped a key safety pledge<\/a> that vowed it would stop training an AI system if it couldn\u2019t guarantee it had proper guardrails in place. And while it fought with the Pentagon against its tech being used in mass surveillance and autonomous weaponry, it emerged that its Claude AI was being used to select strike targets in Iran<\/a>.<\/p>\n One person responding to the company\u2019s ad<\/a> said it best: \u201cCan Anthropic be trusted?\u201d<\/p>\n