Anthropic call for AI pause a 'clever marketing trick' – Irish expert

Anthropic, which owns the Claude AI platform, said R&D into AI should 'temporarily pause' due to fears AI could escape human control. However, Abeba Birhane says their warning is 'misleading and overly exaggerated'. File picture

Tech giant Anthropic’s call for a pause in the development of artificial intelligence (AI) has been dismissed as a “clever marketing trick” by an Irish expert in AI accountability.

Abeba Birhane, the principal investigator with Trinity College Dublin’s artificial intelligence accountability lab, described the warning by Anthropic as “misleading and overly exaggerated”.

The California-based company suggested that research and development into advanced AI should “temporarily pause” due to the increasing possibility that those systems could eventually escape human control entirely.

About 80% of coding and upgrading applied to Anthropic’s Claude platform is being developed by the system itself.

In a blog post, Anthropic suggested that a global pause in development of AI systems would “likely be a good thing”, but that any such initiative would be pointless unless accepted and carried out by all companies and vested interests operating in the industry.

Ms Birhane said there is “no substantial evidence” to support Anthropic’s claims of Claude’s developing autonomy. She said: 

"They are presenting their AI systems as if they are fully autonomous, but this is incorrect,” she said.

“Their top models continually need human verification for fine tuning.

“They are not reliable systems. There are layers of domain experts and contractors that verify code. These systems depend on human interaction.”

Ms Birhane claimed Anthropic’s post represents “a clever kind of framing”.

“They want to abdicate responsibility to their own neural network,” she said. 

It is a clever marketing trick to evade accountability and responsibility for the people designing the system and to give it to a computer network.

They either don’t have the evidence for what they’re saying or they’re not sharing it.”

Anthropic move 'a response to Donald Trump' 

Meanwhile, Ian Dodson, founder and chief executive of Dublin-based education platform AICertified, said Anthropic’s warning may be the company seeking to own its newfound status as the “good” AI company after it fell foul of the Trump administration.

US president Donald Trump in February instructed all US federal agencies to stop using Claude after Anthropic declined to remove safety precautions to allow its systems make autonomous decisions regarding the killing of humans in war scenarios.

It was replaced with Sam Altman’s OpenAI, the firm behind ChatGPT.

Mr Dodson said: “While the interests of corporate shareholders rarely align with those of consumers, in this case Anthropic’s have aligned with ours. 

They’ve gotten their ass kicked by the US government, so now they’re in the anti-Trump camp. 

"They’re saying: ‘We’ve been tagged as the moral guys, so let’s start living up to it.’  

While AI is becoming so developed that it may no longer require humans “to have their hand on the wheel”, Mr Dodson said the outcome is less likely to be apocalyptic.

He pointed to one of the biggest warning signs around AI’s capabilities — that, if the technology continues to develop rapidly, it could pose a threat to modern encryption.

“If you have a system out there with that capability and no human oversight, then every company and nation will want that capability, and it becomes an arms race — like nuclear weapons,” Mr Dodson said.