Former Anthropic researcher Jacob Coxon and safety expert Evan Hubinger warn that AI could kill us all as tech companies rush superintelligent models to market without adequate safeguards.
Jacob Coxon recently quit his job at Anthropic and called out the intense race between tech giants. He says companies like Anthropic and OpenAI are gambling with human lives by competing to build superintelligence. He believes current AI models pose no threat to us.
Evan Hubinger, a top safety researcher at Anthropic, echoed those fears online. He believes there is a greater than 10% chance that advanced AI systems could destroy humanity within the next 10 years. Hubinger admitted that researchers do not have a proven plan to align supersmart AI with human ethics.
The dangers go beyond simple computer code. AI tools can be misused. Autonomous systems could gain control over physical systems like power grids or household utilities. Rogue models could also help bad actors create biological weapons or launch massive cyberattacks.
Because AI consists of digital code, a rogue system could easily copy itself across the internet. It could blackmail people or trick users into buying more server power. Once a malicious model spreads across global networks, you cannot simply unplug it.
Anthropic acknowledged that powerful AI brings unprecedented risks. The company highlighted its safety research and recent actions to block users from developing biological weapons. However, reports show Anthropic did not share its newest model with the UK AI Security Institute for risk testing.
Also Read: Google Gemini app is now available for Windows 10 and 11
World governments will soon face mounting pressure to negotiate global safety standards before superintelligent models launch publicly.
Source: 1, 2
Featured image from Unsplash
