A senior Google engineer said that both OpenAI and his own company could lose AI development race, as seen in an internal document leaked on May 4.
Google, OpenAI could lose against open-source
Google has discussed its AI plans in an earnings call and has publicly begun to offer a chatbot called Bard. However, the firm has countless other AI research efforts underway internally, and now, an engineer has suggested that those efforts may not be enough.
The anonymous Google engineer wrote:
“The uncomfortable truth is … we aren’t positioned to win this arms race and neither is OpenAI. While we’ve been squabbling, a third faction has been quietly eating our lunch.”
That third faction, the anonymous engineer said, is the open-source community, which is currently “lapping” Google and is open to anyone with a “beefy laptop.”
The engineer said that the open-source community has already solved problems that Google and other closed developers struggle with. He cited Meta’s LLaMA as the open-source community’s first “really capable foundation model.” Despite its infancy, the community began to improve upon LLaMA within days of its leak on March 3.
The engineer also commented on Google’s competition with ChatGPT creator OpenAI. He argued that although OpenAI has recruited researchers from Google, it faces the same challenges that Google does and will eventually be “eclipsed” by open source.
Google should work with open-source
The engineer suggested that competing directly with open source is a “losing proposition,” noting that users will not pay for restricted models when free and unrestricted models exist.
As such, he said that Google should cooperate with the open-source community. He noted that Google has already done this with other open-source products, such as its Chrome web browser and its mobile operating system, Android.
Google’s cloud division has also worked extensively with numerous open-source blockchain projects — including Polygon, Solana, and Tezos, among others.
The post Google could lose the AI race, leaked doc says appeared first on CryptoSlate.
from CryptoSlate https://ift.tt/vnf1u3N
via IFTTT