Recent headlines from tech giants Google and Meta about artificial intelligence (AI) have added fuel to the debate around open-source development. On Tuesday evening, CNBC reported that Google’s language model (PaLM 2) uses five times more data for training than its predecessor but the company has been unwilling to provide further details. In response, Google DeepMind software engineer Dmitry Lepikhin tweeted “whoever leaked PaLM2 details to cnbc, sincerely fuck you!”
At the same time, Alex Polozov, a senior staff research scientist at Google, pointed out that this will lead to increased secrecy when it comes to AI research, while Google AI researcher Lucas Beyer expressed concern about a “worse work/research environment.”
In a coincidental timing, Meta chief AI scientist Yann LeCun spoke to the New York Times about his company’s open-source AI efforts. In February, Meta released its large language model (LLaMA) and made the source code available to academics and government researchers who gave their email address. LeCun’s position is that the platform that will succeed is the open one, and the increase in corporate secrecy at Google and OpenAI is a “huge mistake” and a “really bad take”.
Meta’s Yann LeCun has also had a hand in open source development as he was one of the creators of the PyTorch platform that is used by OpenAI. Nevertheless, even Meta and LeCun go only so far when it comes to openness and, in Yann’s words, it is not always “the responsible thing to do”. Furthermore, he argues that the most dangerous element is the dissemination of misinformation on social media, something for which there is no AI technology to counter.
Google and OpenAI have become more secretive about their AI research, but Meta and its chief AI scientist Yann LeCun still advocates for the open-source development of AI. He believes that allowing access and transparency will help with verifiability, while stressing that some level of openness should still be retained. Furthermore, Yann warns against the most dangerous element of AI: the potential to spread misinformation on social media.
