Meta AI hosted an event in London on Tuesday where discussions revolved around topics such as the democratization of artificial intelligence, AI’s potential threats to democracy, and the upcoming release of Meta’s Llama 3.
Nick Clegg, former UK Deputy Prime Minister and current President of Global Affairs at Meta (formerly known as Facebook), and Yann LeCun, Meta’s chief AI scientist, argued that artificial intelligence (AI) needs to be liberated from the monopolistic control of a few large US tech companies, to ensure that it doesn’t remain concentrated in the “clammy hands” of Silicon Valley.
Amplifying Meta’s business ethos, Clegg emphasized that AI tools should be made widely and freely available, a view that LeCun ardently seconds. They both believe in decentralizing AI towards a more open-source approach—even though Meta’s Llama series of language models are not entirely open-source, they are somewhat more democratized than those offered by the likes of OpenAI or Google.
The democratization of AI is also a topic that has been previously broached by NVIDIA CEO Jensen Huang and Emad Mostaque, formerly of Stability AI, who both underlined the need for nations to create their own sovereign AI, thereby removing the technology from its current centralized ownership.
Clegg also went against the grain by underscoring that AI tools haven’t yet been used systematically to disrupt or subvert major elections in countries like Taiwan, Pakistan, Bangladesh, and Indonesia. This was inspite of well-documented instances in these countries where AI-enhanced deepfakes were used to disseminate political misinformation, with the potential to influence voter sentiment.
Clegg was not marginalizing the potential harms of AI, but was advocating for a balanced view of AI as both a defensive and offensive tool against disinformation—our “sword and shield” against false information, according to him.
The event also shed light on Meta’s upcoming plans—the deployment of their next-generation language model, Llama 3. Although more specific details were withheld, Meta confirmed that the new model would be more open with weaker or flexible guardrails, and its release would mark yet another step towards their commitment to a more democratized AI. With around 140 billion parameters, Llama 3 is expected to be far more advanced than its predecessor, Llama 2, which had 70 billion parameters.
In conclusion, the Meta AI event touched upon key issues and trends in the AI space, including the democratization of AI, AI’s role in politics, and Meta’s own ambitions for future development with the unveiling of Llama 3.
