Искусственный General Int ...

Каким образом непрофессионал может отличить обман искусственного интеллекта от реального технического прогресса?

Любит | Нелюбов | Ответы | Вид | 6215

Шрираман Мадхаван Стэнфордская Статистика |   
@ | Обновлено Right Now

  • Any sentence which talks about how “smart” AI is, without mentioning a particular task, is definitely hype. AI is not an entity. All of AI progress is task-specific: Object detection, Music generation, Image classification, playing Go, etc. Saying AI is smarter than a human is meaningless. An AI system can perform a particular task for which it was built, better than a human.
  • Any sentence which contains both “AI” and “sentient” is probably hype. Unless the sentence is “Any sentence which contains both “AI” and “sentient” is probably hype.” :P We wouldn’t say logistic regression is self-aware, would we? Today’s AI systems are not that far off.
  • Any sentence which compares today’s AI systems with Hollywood AI is definitely hype. There is no self-evident path from today’s AI to AGI (Artificial General Intelligence - the kind you see in Hollywood and sci-fi novels.) It’s like claiming we are closer to the moon after climbing a tree.
  • Any sentence which compares AI systems to how the human brain works is probably hype. Of course, borrowed terminologies like “neurons”, “activations”, etc. certainly do not help, but AI systems do not mimic the human brain in any way except in structure. That’s mostly because little is known about how the human brain works in the first place.
  • Any sentence which anthropomorphizes AI i.e., attributes any lesser-understood human qualities like creativity, empathy, consciousness, etc. to AI is definitely hype. Eg.: “Google AI creates child AI”, “Facebook shuts down AI bots after they invent language”, etc.

Half-baked “news” articles with click-***** headlines aren’t entirely to blame for the AI hype-reality rift we see today. Even certain corporations publicize outlandish claims that aren’t backed by credible data. There’s a reason Oren Etzioni (CEO of the Allen Institute for AI) called IBM Watson “the Donald Trump of the AI industry.” Hanson Robotics’ Sophia is another example.

As an antidote, I would recommend following people like Andrew Ng, Yann LeCun and Geoff Hinton. Be very skeptical of AI-related news articles. Scrutinize every sentence, every quote and every person quoted. If something sounds too good to be true, it probably is.

Yes, I’m fun at parties.

| |

«Машинное обучение для людей»
Один из самых трендовых курсов в сфере Цифровой экономики.