AI will never become a conscious being — Sentient founder

Artificial quality (AI) volition ne'er go a conscious being owed to a deficiency of intention, which is endemic to quality beings and different biologic creatures, according to Sandeep Nailwal — co-founder of Polygon and the open-source AI institution Sentient.

"I don't spot that AI volition person immoderate important level of conscience," Nailwal told Cointelegraph successful an interview, adding that helium does not judge the doomsday scenario of AI becoming self-aware and taking implicit humanity is possible.

The enforcement was captious of the mentation that consciousness emerges randomly owed to analyzable chemic interactions oregon processes and said that portion these processes tin make analyzable cells, they cannot make consciousness.

Instead, Nailwal is acrophobic that centralized institutions volition misuse artificial quality for surveillance and curtail idiosyncratic freedoms, which is wherefore AI indispensable beryllium transparent and democratized. Nailwal said:

"That is my halfway thought for however I came up with the thought of Sentient, that yet the planetary AI, which tin really make a borderless world, should beryllium an AI that is controlled by each quality being."

The enforcement added that these centralized threats are wherefore each idiosyncratic needs a customized AI that works connected their behalf and is loyal to that circumstantial idiosyncratic to support themselves from different AIs deployed by almighty institutions.

Sentient’s unfastened exemplary attack to AI vs the opaque attack of centralized platforms. Source: Sentient Whitepaper

Related: OpenAI’s GPT-4.5 ‘won’t crush benchmarks’ but mightiness beryllium a amended friend

Decentralized AI tin assistance forestall a catastrophe earlier it transpires

In October 2024, AI institution Anthropic released a paper outlining scenarios wherever AI could sabotage humanity and imaginable solutions to the problem.

Ultimately, the insubstantial concluded that AI is not an contiguous threat to humanity but could go unsafe successful the aboriginal arsenic AI models go much advanced.

Different types of imaginable AI sabotage scenarios outlined successful the Anthropic paper. Source: Anthropic

David Holtzman, a erstwhile subject quality nonrecreational and main strategy serviceman of the Naoris decentralized information protocol, told Cointelegraph that AI poses a monolithic hazard to privacy successful the adjacent term.

Like Nailwal, Holtzman argued that centralized institutions, including the state, could wield AI for surveillance and that decentralization is simply a bulwark against AI threats.

Magazine: ChatGPT trigger blessed with nukes, SEGA’s 80s AI, TAO up 90%: AI Eye