• Near Human Abilities

    From Mike Powell@1:2320/105 to All on Tue Mar 5 10:20:00 2024
    * Originally in: Dovenet TechTalk

    From TLDR 2024-03-05

    THE AI WARS HEAT UP WITH CLAUDE 3, CLAIMED TO HAVE "NEAR-HUMAN"
    ABILITIES (4 MINUTE READ) [6]

    Anthropic's Claude 3 apparently sets new industry benchmarks across a
    range of cognitive tasks. The company claims that it approaches
    'near-human' capability in some cases. Claude 3 has three models:
    Claude 3 Haiku, Claude 3 Sonnet (which powers the Claude.ai chatbot),
    and Claude 3 Opus (available through a subscription service). All
    three models feature a 200,000-token context window and are available
    through an API. Claude 3 beats GPT-4 on a range of widely used
    benchmarks.

    AGGREGATOR'S AI RISK (27 MINUTE READ) [12]

    The Internet made publishing and distribution zero marginal cost
    activities. It created a new world of abundance, but it also created a
    new type of power given to those who had the ability to make sense of
    that abundance. Allowing anyone to publish didn't further decentralize
    power - new powers emerged to control discovery. Generative AI flips
    that paradigm around, collapsing all published knowledge into a single
    answer. This answer is unlikely to make everyone happy. A potential
    solution to this problem is to personalize AIs so they do a better job
    of giving people what they want to see.

    Links:
    ------
    [6] https://arstechnica.com/information-technology/2024/03/the-ai-wars-heat-up-with-claude-3-claimed-to-have-near-human-abilities/?utm_source=tldrnewsletter

    [12]
    ttps://stratechery.com/2024/aggregators-ai-risk/?utm_source=tldrnewsletter

    -- https://refer.tldr.tech/1fd848fa/

    $$
    --- SBBSecho 3.20-Linux
    * Origin: capitolcityonline.net * Telnet/SSH:2022/HTTP (1:2320/105)