hn.today

So Far, AI Pessimism Is Observationally Equivalent to AI Optimism

betonit.ai7 points5 comments
Screenshot of So Far, AI Pessimism Is Observationally Equivalent to AI Optimism

Bryan Caplan argues that AI pessimists deserve credit for predicting rapid progress: AI has exceeded even his expectations, and he considers it broadly capable enough to count as AGI. But optimists also anticipated that progress when AI was still weak, so both camps have a record of foresight and current evidence does not settle whether AI will ultimately be disastrous or beneficial. He argues that scandals and persistent harms do not disprove optimism, since beneficial technologies also produce serious problems. Cars kill more than a million people annually without prompting calls to halt their use, while fear of nuclear accidents helped suppress a technology he regards as exceptionally safe, clean, and cheap. In his view, AI’s potential benefits justify tolerating substantial collateral damage; pessimists need evidence of catastrophe, not merely alarming incidents.

The disagreement, he says, turns mainly on prior beliefs and the historical comparisons used to make predictions. Analogies can support almost any conclusion, but good forecasters still choose useful base rates. Caplan favors treating predictions of imminent human extinction as belonging to the record of past extinction forecasts, which have failed, rather than assuming that AI’s growing power makes those warnings credible. Common sense, he argues, supports the view that technological progress is beneficial overall, even though particular breakthroughs are hard to predict and generate complaints. He rejects extravagant optimism about a singularity or explosive economic growth, but concludes that optimists have the stronger case: AI’s rapid advance is established, and its future may be bright despite ongoing problems.

Read on betonit.ai5 comments on Hacker News

Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.

More in AI

The daily digest

Today's best Hacker News stories, summarized and screenshotted, one email a day.