Elon Musk says AI has 'more ¹ÎÄêŪ¤Ês than ¾Ã¶ËŪ¤Ês' as he ³«»Ï¤¹¤ë¡¤ÂǤÁ¾å¤²¤ës his own anti-woke »¨Ã̡ʤ¹¤ë¡Ë bot - in first public ³°¸« since ²ò¸Û¤¹¤ë¡¿Ë¤²Ð¡¿¼Í·âing Don Lemon
- Musk ¤ËÎ㤨¤ëd dangerous AI to HAL in Arthur C. Clarke's '2001: A Space Ĺ´üËÁ¸±Î¹¹Ô'
- READ MORE: ÃÏ¿Þ¡¿·×²è¤¹¤ë ÌÀ¤é¤«¤Ë¤¹¤ë¡¿Ï³¤é¤¹s US cities?DOOMED to lose the most ¿¦¶Ès to AI tech
AI doomster Elon Musk appears to have changed his tune on the ²Ê³Ø¡Ê¹©³Ø¡Ëµ»½Ñ, ¡Ê¿ÍÌ¿¤Ê¤É¤ò¡ËÃ¥¤¦¡¤¼çÄ¥¤¹¤ëing the ¥×¥í¤Î¡¿»¿À®¤Îs of Á°¿Ê¤¹¤ëing the tech ºÇ½ªÅª¤Ë outweigh the È¿ÂС¿º¾µ½s.
Musk was speaking?¸øÁ³¤È for the first time at an ¿Í¹©Åª¤Ê ÃÎǽ ¼óǾ²ñµÄ Tuesday, a little ¤¤¤Ã¤½¤¦¾¯¤Ê¤¯ than a week after the ²¯ËüĹ¼Ô ²ò¸Û¤¹¤ë¡¿Ë¤²Ð¡¿¼Í·âd?Don Lemon.
The ̵½ê°¤ÎÀ¯¼£²È tech mogul?¡½ who °ÊÁ° called AI the 'most destructive ·³Ââ in history' ¡½ »ý³¤¹¤ëd there was still 'some chance that it will end humanity.' But his ȯ¼Í¡¿¿äÄês were ¤¤¤Ã¤½¤¦¾¯¤Ê¤¯ °Å¤¤¡¿Í¥¤¦¤Ä¤Ê than usual.?
'I probably agree with Geof Hinton that it's about 10 ¥Ñ¡¼¥»¥ó¥È or 20 ¥Ñ¡¼¥»¥ó¥È,' Musk told ¼óǾ²ñµÄ attendees, namechecking?Geoffrey Hinton, Google's?'Godfather of AI.'??
'But I think we are Ĺ¡¤Î¨¤¤¤ëd for ËÉÙ,' Musk opined, 'as the most likely ·ë²Ì.'
'I think the probable ¹ÎÄêŪ¤Ê ¥·¥Ê¥ê¥ª outweighs the ¾Ã¶ËŪ¤Ê ¥·¥Ê¥ê¥ª,' he said.
The ÌÀ¤é¤«¤Ê ¥È¥óing É餫¤¹¡¿·âÄƤ¹¤ë of Musk's ¶²¤ì¤ës about AI come as he ³«»Ï¤¹¤ë¡¤ÂǤÁ¾å¤²¤ës his own ¿Í¹©Åª¤Ê ÃÎǽ program »Ï¤á¤ë¡¤·è¤á¤ë to compete with ChatGPT À½Â¤¼Ô OpenAI: a 'sarcastic' AI dubbed Grok that's now in beta on Musk's rebrand of Twitter, X.?
AI doomster Elon Musk (above Ãæ¿´) spoke at an ¿Í¹©Åª¤Ê ÃÎǽ ¼óǾ²ñµÄ Tuesday, Abundance360, in conversation with the group's co-ÁÏΩ¼Ô Peter Diamandis (¸¢Íø). Musk told attendees 'the probable ¹ÎÄêŪ¤Ê ¥·¥Ê¥ê¥ª' for AI 'outweighs the ¾Ã¶ËŪ¤Ê ¥·¥Ê¥ê¥ª'
Hinton, 75, a British-Canadian AI ³«Â󤹤ë who?sensationally ¤ä¤á¤ë his ¿¦¶È at Google last May?over his own ¶²¤ì¤ës and ²ù¤¤¤ës about AI, also spoke at this week's ¼óǾ²ñµÄ.
The event, Abundance360?by Singularity University, was created by futurist and X Prize ÁÏΩ¼Ô Peter Diamandis, who has ³ôd Musk's passion for the »äŪ¤Ê space race, boot-strapping a variety of start-ups ¸¥¿ÈŪ¤Ê to micro-±ÒÀ±s, space ´Ñ¸÷ι¹Ô¡¤»ö¶È and even lunar rovers.
Diamandis interviewed Musk for Tuesday's live event ·Ðͳ¤Ç a ¤¤¤Ä¤«s imperfect Starlink-link supported, X.com ¥Ó¥Ç¥ª livestream, as Musk sat on a ·×²è¡Ê¤¹¤ë¡Ë bound for Los Angeles.
Geoffrey Hinton, 75?¡ ½ a British-Canadian AI ³«Â󤹤ë who ÌäÂ꡿ȯ¹Ô¤¹¤ëd a ·Ù¹ð that 'scary' chatbots like the popular ChatGPT could soon be smarter than humans?¡½ also spoke at this week's Abundance360 ¼óǾ²ñµÄ
Branded as the '¹Âç¤Ê¡¿Â¿¿ô¤Î¡¿½ÅÍ×¤Ê AI ¿³µÄ,' this year's ¸«²ò¡¿ËÝÌõ¡¿ÈÇ of Diamandis' event spanned four days in?Palos Verdes, California. ¡Ê½°µÄ±¡¤Î¡ËµÄĹ's ´Þ¤àd?Google futurist and coiner of the ¾Î¤¹¤ë¡¤¸Æ¤Ö¡¿´ü´Ö¡¿ÍѸì 'the Singularity,'?Ray Kurzweil, and?Trump-ÍƼÏd, 1980s 'junk ¼ÒºÄ' financier?Michael Milken.
Former Google CEO?Eric Schmidt, who said last year that AI Ä󵯤¹¤ë¡¿¥Ý¡¼¥º¤ò¤È¤ëd an?'existential ´í¸±' for humanity 'defined as many, many, many, many people ³²¡Ê¤òÍ¿¤¨¤ë¡Ëd or killed,' also took part in the ¼óǾ²ñµÄ.?
'It's called 'the singularity' for a ¿äÏÀ¤¹¤ë¡¿Íýͳ,' Musk quipped, ¸ÀµÚ¡¿´ØÏ¢ing Kurzweil's theory that a ̤Íè moment of exponential ²Ê³Øµ»½Ñ¤Î ¿ÊÊâ will forever ¹çÊ»¤¹¤ë humanity with machines into a brilliant new Ãʳ¬ of ¿Ê²½.?
'When you have the advent of ¿Í¹©Åª¤Ê superintelligence,' Musk continued, 'it is ¸½¼Â¤Ë very difficult to ͽÊ󤹤ë what will happen next.'
But, he Äɲ乤ëd: 'It's definitely going to happen ¡½ and it's happening µÞ®¤Ê¡¿ÊüÆ¢¤Ê. So, I think really we just want to steer it in as ¹ÎÄêŪ¤Ê a direction as possible.'?
The space-tech galvanizing, X Prize ÁÏΩ¼Ô Peter Diamandis (¹Ô¤¦¡¿³«ºÅ¤¹¤ë¡¿Ãʳ¬ left)?interviewed Musk for Tuesday's live event ·Ðͳ¤Ç a ¤¤¤Ä¤«s imperfect Starlink-link supported, X.com broadcast livestream, as Musk spoke while seated on a ·×²è¡Ê¤¹¤ë¡Ë bound for Los Angeles
The 'godfather' of AI Geoffrey Hinton (Ãæ¿´ ¸¢Íø) Åꤲ¾å¤²¤ë¡¿¥Ü¥Ç¥£¥Á¥§¥Ã¥¯¤¹¤ëd a ¼ê¤ê¤å¤¦ÃÆ into the ·ãÅܡʤ¹¤ë¡Ëing ¿³µÄ about the dangers of AI ²Ê³Ø¡Ê¹©³Ø¡Ëµ»½Ñ last May, after sensationally quitting his ¿¦¶È at Google. Tuesday, Musk said he agreed with Hinton, while still »ö¶È¡¿·×²è¡Ê¤¹¤ë¡Ëing a more rosy »ëÌî on AI
Musk drew ½ª»Ï°ì´Ó¤·¤Æ from the realm of science fiction in a search for metaphors to better explain his ¸«Ä̤· on the ²ÄǽÀ¤Î¤¢¤ë Ìó« and Í·ês of ¹âÅÙ¤Ë Á°¿Ê¤¹¤ëd, ¿Í¹©Åª¤Ê general ÃÎǽ (AGI) and other more ÌÀ³Î¤Ê¡¿ºÙÉô forms of AI.
'I think we'll have an ·ë²Ì that's Îà»÷¤Î to the Ian Banks' "Culture" Ä´½ñ¤ò¤È¤ë¡¿Í½Ì󤹤ës,' Musk said from his airplane seat, calling the late Scottish ¾®Àâ²È Banks' °ìÏ¢¤Î ten Culture novels ¤«¤â¤·¤ì¤Ê¤¤ 'the best ÁÛÁü¤¹¤ëing of a Ⱦʬ-utopian AI ̤Íè.'
Banks' Culture novels take place in a Æý¤Î Way À±±À ¤À¤¤¤¿¤¤ 9,000 years after humanity and other human-like ¼ïÎà have joined with AI to produce a 'ÃÏ°Ì¡¤Ç¤Ì¿¤¹¤ë-scarcity' economy with little need for ˡΧs or ˡΧ »Ü¹Ô.
Banks once ½Ò¤Ù¤ëd the utopian ¸«Ä̤· of his Ä´½ñ¤ò¤È¤ë¡¿Í½Ì󤹤ës as 'space ¼Ò²ñ¼çµÁ.'
Musk then pointed to the »¦¤·²° robot HAL from Arthur C. Clarke.'s '2001: A Space Ĺ´üËÁ¸±Î¹¹Ô' as an example of how and why AI could become a ¶¼¤· to humanity.
The '³Ë¿´ Á°Äó' of 2001, as Musk put it to the ¼óǾ²ñµÄ, was that 'things went wrong when they ·³Ââd the AI to ±³¡Ê¤ò¤Ä¤¯¡Ë.'?
'You kinda do grow an AGI,' Musk said.?
'It's almost like raising a kid, but a ºÇ¹â¤Î-genius, God-like ÃÎǽ kid ¡½ and it »öÊÁs how you raise the kid, you know?'
In keeping with his past comments on the ÌäÂ꡿ȯ¹Ô¤¹¤ë, Musk »ý³¤¹¤ëd that the safest AI for humanity is one that is 'g rown' to remain?'maximumly truth-Áܤ·½Ð¤¹ing and curious.'
'The best way to ãÀ®¤¹¤ë AI safety is to just grow the AI, ¡¼¤Ë´Ø¤·¤Æ¡¿¡¼¤ÎÅÀ¤Çs of the ÁÏΩ¡¿´ðÁà model and the ȳ¶â-tuning, to be really truthful ¡½ like don't ·³Ââ it to ±³¡Ê¤ò¤Ä¤¯¡Ë, even if the truth is unpleasant,' Musk said.
The ²¯ËüĹ¼Ô, who ¹ØÆþ¡Ê¤¹¤ë¡Ëd the social ¥Þ¥¹¥³¥ß ¾ì½ê¡¿°ÌÃÖ Twitter in October 2022, then put in a plug for his ¾ì½ê¡¿°ÌÃÖ, now called X, and its own AI chatbot in beta, Grok.?
Grok, Musk said to a wave of ¾Þ»¿ from the audience, will be 'maximally truthful' even if what it says is not 'À¯¼£¾å ÄûÀµ¤¹¤ë.'
?Grok, 'a humble AI assistant,' is only ÍøÍѤǤ¤ë to X ¾ÞÍ¿¶â+ ²ÃÆþ¼Ôs.