An artificial-intelligence launch can be a risky business these days, such is the competition in Big Tech and the hype among investors.
But Google’s release of Gemini—its most advanced AI software model—is being well received. Shares of
Alphabet,
Google’s parent, climbed 4.8% to $136.25 early Thursday.
That will be a relief for the company, which is all too aware that AI product announcements can be a potential banana skin.
Alphabet
lost $100 billion in market value in one day after an event to promote its chatbot Bard in February. The event, rushed into the calendar ahead of
Microsoft’s
Bing Chat announcement, left investors uninspired.
The Bard launch is looking more like a one-off, though, and Google is now very much a major player in the global AI race.
J.P.Morgan analyst Doug Anmuth wrote that the Gemini release Wednesday represents “significant innovation for Google.” Anmuth reiterated his Overweight rating on Alphabet stock, with a target price of $150.
“While the Street mostly yawned at the release yesterday—perhaps viewing the product as not quite complete and the timing both pressured and opportunistic—we are encouraged to see Google’s progress on this major technology shift,” he wrote.
Anmuth noted that Gemini scored higher than OpenAI’s GPT-4 on the industry-standard MMLU, or massive multitask language understanding, benchmark. It is also the first model to outperform human experts. MMLU addresses knowledge in a combination of 57 subjects, including math, physics, history, law, medicine, and ethics, among others.
Google is launching three versions of Gemini for various applications.
The most powerful version, Gemini Ultra, is still being safety tested and reviewed, with an expected launch date earlier next year. Midline version Gemini Pro, the company said, will scale across many tasks, and is already leveraged for some answers in Bard. Gemini Nano is the smallest version of the model, intended for use on smartphones, laptops, and other local devices.
Write to Callum Keown at callum.keown@barrons.com
Read the full article here