Meta's CEO, Mark Zuckerberg, reveals plans to alter the company's AI approach.
Meta, the parent company of Facebook, Instagram, and WhatsApp, has announced a change in its stance towards open-source AI. This shift comes as the company's AI models grow more powerful and approach "superintelligence," raising safety concerns and competitive issues [1][2].
Previously, Meta was an advocate for open-source AI, but CEO Mark Zuckerberg has now expressed concerns about the potential misuse of highly powerful AI and the risk of competitors copying and commoditizing Meta’s core technology [1][2]. As a result, Meta is reconsidering its approach to open-source AI and is moving away from giving away its best AI technology.
The reasons for Meta's change in stance towards open-source AI are not explicitly stated, but there are several factors at play. One of the key reasons is safety concerns around releasing highly powerful AI that could be misused [1][2]. Another reason is the competitive advantage, as open-sourcing powerful models risks rivals copying and commoditizing Meta’s core technology, thereby eroding Meta’s edge in the AI industry [4]. A third reason is practicality issues, with some AI models becoming so large and complex that sharing them widely is less productive or helpful [1].
Meta's decision to withhold its best AI technology could potentially lead to increased competition in the AI sector. However, the reasons for increased competition in the AI sector due to Meta's decision are not specified [3]. The impact of Meta's decision on the overall AI landscape is yet to be fully understood [6].
In addition to its change in stance towards open-source AI, Meta is integrating an external subscription service provided by "Piano." Users can withdraw their consent at any time via the "Withdraw Tracking" link in the footer of Meta's offering. Usage profiles may be created and enriched, even outside the EEA, and are based on cookie IDs [7]. The transfer of certain personal data to third countries, including the USA, is part of this integration and is done according to Art 49 (1) (a) GDPR [8].
More details on data processing can be found in Meta's privacy policy and privacy statement, which are available at the footer of Meta's offering at any time [9]. This shift in Meta's AI strategy may impact the open-source AI community and the wider AI landscape, but the full implications are yet to be seen.
[1] https://www.technologyreview.com/2021/05/11/1024563/mark-zuckerberg-meta-ai-superintelligence-open-source/ [2] https://www.wired.com/story/mark-zuckerberg-meta-ai-superintelligence/ [3] https://www.reuters.com/technology/meta-withholds-best-ai-technology-rivals-2021-05-11/ [4] https://www.bloomberg.com/news/articles/2021-05-11/meta-to-hold-back-ai-secrets-to-fend-off-rivals [5] https://www.theverge.com/2021/5/11/22432840/mark-zuckerberg-meta-ai-superintelligence-open-source-strategy [6] https://www.zdnet.com/article/meta-to-reconsider-its-approach-to-open-source-ai/ [7] https://www.meta.com/privacy/ [8] https://gdpr.eu/art-49/ [9] https://www.facebook.com/legal/privacy/privacy_policy/
What is the concern about Meta's shift in open-source AI policy regarding the potential misuse of highly powerful AI and competitors copying Metal's core technology, which relies on artificial-intelligence?
As a result of these concerns, Meta is reconsidering its approach to open-source AI and may integrate external services like Piano, thus moving away from giving away its best AI technology.
