"Dem go soon get sense pass us" - AI 'godfather' warn about dia dangers as e resign from Google

One man wey dem dey see as di godfather of artificial intelligence (AI) don resign for im job as im warn about di kain kasala wey fit burst sake of di development for di field.

Geoffrey Hinton wey be 75 years old announce im resignation from Google for statement wey im give di New York Times say e don dey regret im work now.

E tell BBC say some of di dangers of AI chatbox dey "quiet scary".

E say, "right now, dem neva dey intelligent pass us as I fit see. But I tink say soon, dem fit be".

Dr Hinton also accept say true-true, im age fit be part of di reason why im decide to leave Google as im tell BBC say, "I be 75, so time don reach to retire."

Dr Hinton ogbonge research for deep learning and neural networks na im clear road for di current AI systems like ChatGPT.

But di British-Canadian cognitive psychologist and computer scientist tell BBC sat di chatbox fit overtake di level of information wey di human brain dey hold.

"Right now, wetin we dey see na tins like GPT-4 wey dey pass pesin in di amount of general knowledge wey im get and do pass dem by far. For reasoning, dem neva dey dat good but dem fit already dey do simple reasoning", na wetin e tok.

Dr Hinton add say, "and given di rate of progress, we dey expect say tins go beta quick-quick. So we gatz worry about dat."

For di New York Times article, Dr Hinton tok about "bad actors" wey fit use AI do "bad tins".

Wen BBC say make e explain am more, e say, "dis na just worst case scenario, kind of nightmare scenario.

"You fit imagine, for instance, say one bad actor like [Russia President Vladimir Putin] decide to give robots di ability to create dia sub goals."

Di scientist warn say e fit make dem "create sub-goals like "I need to get more power".

E add say, "I don reach conclusion sat di kain intelligence we dey develop dey very different from di intelligence wey we get.

"We be biological systems and dis na digital systems. And di difference be say with digital systems, you get many copies of di same set of weights, di same models all ova di world.

"And all dis copies fit learn separately and share dia knowlegde instantly. Na like if you get 10,000 pipo and wen one pesin learn sometin, everybodi go just know am instanta. Na how dis chatbots fit sabi pass any one pesin."

E stand strong say e no wan yab Google say di tech giant don dey "very responsible".

E say "I wan say beta tins about Google. And e go dey more credible if I no dey work for Google".

For statement, Google chief scientist Jeff Dean say, "we dey committed to responsible approach to AI. We dey continue to learn to understand di upcoming risks as we dey innovate boldly."