Forging Ahead: AI Giants Unite for Frontier Model Forum on AI Safety

0:00The AI tech giants have made a major move. And this comes on the back of a recent visit to the White House where all the major tech firms essentially pledged to focus more on tech and AI safety with these tech models, with these AI models. And so today there is a new forum that has actually been launched with industry leaders, Anthropic, Google, Microsoft, and OpenAI. They've all essentially come together to launch what is called the Frontier Model Forum. So today on the podcast, we're going to be diving into what exactly this is and where we see the implications of this going in the future. how we see this really shifting the landscape of AI, and perhaps some sort of consolidation of maybe not power, but consensus among the top AI leaders.

0:43So let's dive into it. So this new venture aims to really ensure the responsible and safe development of advanced AI models, which have started to kind of blur the boundaries between human and machine capabilities. Frontier models, by definition, are large-scale machine learning models whose capabilities essentially surpass those of current cutting-edge models, enabling them to undertake a really broad range of tasks. So recognizing the potential risks associated with these, you know, quote-unquote frontier models, this new frontier model forum is going to focus on establishing safety standards, identifying best practices, and also encouraging research and development towards mitigating potential threats.

1:22I believe that with, you know, the collective expertise from some of these big tech companies that form is really kind of trying to advance AI safety research. They're trying to support applications that address societal challenges like cybersecurity. And they're also trying to foster knowledge sharing with policymakers, academics, and the public. The ultimate goal is to create a transparent, informed environment for the development and deployment of Frontier AI models. Now, will that be possible? Are they going to achieve this? I mean, I wouldn't 100 % say yes to it but I think it is a promising step that they are moving in this direction that they're thinking about this do I like the consolidation of power it's kind of interesting because like on the one hand um you see so I'll tell you the criticism that people are going to give to this number one you kind of saw what was called the light bulb cabal or something where essentially phillips light bulb I believe don't sue me phillips I just go do your own research on the story but essentially there's a bunch of light bulb companies they all got together and they decided that for the safety of life bulbs they were only going to make them so the life bulbs instead of lasting 100 years only lasted two years or five years or whatever and if you look at the lifespan of your average life or light bulb it has continuously declined as this group of light bulb makers have gotten together to determine light bulb safety or whatever they decided to deem it as and a lot of people criticize this as look you're literally just all kind of colluding together to do one thing that's better for you and worse for the industry right if light bulbs don't last as long um then you got to buy them more often i believe there's some light bulbs in like philadelphia or somewhere in like a fire department um that have been going for like almost 100 years or something crazy so light bulbs really do have the capability of lasting for a very long time but the current light bulbs we have today do not do that in fact i sometimes get frustrated because i'll buy these like 20 year light bulbs from walmart and plug them in and like one year to two years later they're all burnt out and as if i'm gonna go take the light bulb back to Walmart and get my the four dollars back I don't have the original packaging I don't know where the receipt is yada yada yada so I just think um that's the criticism people are going to give to this it's going to be the same thing they're going to be worried that when you have the big players like Google and OpenAI and Microsoft all together in this big AI forum that perhaps there's going to be a little bit of colluding and in addition to this the forum they get to decide who they allow into the forum so it's kind of like a club and they get to decide who comes into it and maybe you know anthropic has a competitor they don't like and they decide that they don't get to join the club and then if you're not certified um you know a good player in the frontier model forum maybe you don't get to get certain contracts with companies or the government or other things so there's kind of this whole insider trading feeling that i know people will push back against but at the same time um i think it is you know important sometimes for industry leaders in these big firms to come together and make consensuses on things you see this even happening in other industries for example the electric car market right there was all these different um charging ports random example there's all these different charging ports and tesla came out with the north american standard um electric charging port and in order to make it so that everyone when they go to an ev charging station is using the same adapter they don't have to get all these different adapters and like maybe some you know ford can't go to Tesla super chargers because they don't have the right adapter and you know to avoid problems like that essentially you had all of the different car companies in North America kind of coming together and adopting this uniform you know the North American charging standard for EV cars which was essentially the Tesla adapter and I think they just open source to the the patents or let everyone use the patents to to use this specific type of adapter and now it's pretty much unlocked the thousands of ai or the thousands of tesla superchargers across the country and around the world um to every car owner that will eventually right we had ford general motors rivian we've had a whole bunch of um different car companies say that they're going to be adopting this you know this uh standard charging port and i think it makes a lot of sense right they can tap straight into infrastructure that's already built we don't have to build different charging ports for different types of cars it would have been a big hassle so something like that you see like oh man there's some serious benefit to the consumer and to everyone when these big companies come together and come up with a consensus on some things, some standards and whatnot, and best practices.

5:43So that is, I think, a pro to this, but then there is also the cons. So I think there's going to be two sides to this, and we're really going to have to watch this to see how this plays out. But essentially, the Frontier Model Forum is going to structure itself around an advisory board to guide its strategy and priorities. And they say that they're going to be leveraging diverse backgrounds and perspectives the forum also encourages participation from other organizations developing frontier ai models fostering collective approaches towards the safe progression of these models that's kind of their idea here but in the coming years the forum is primarily going to focus on three areas so number one establishing best practices advancing ai safety research and then finally facilitating information sharing among companies and governments, which of course is where people start to sign some, see some alarm bells is the whole information sharing among companies and governments.

6:37It just feels like a little bit less privacy. So some people definitely don't like that. I think this will involve promoting knowledge sharing across different sectors, identifying critical research questions in AI safety and establishing secure mechanisms for sharing information regarding AI safety and risks. leading figures from the founding companies have actually expressed like they're pretty optimistic about this whole initiative so Kent Walker he's the president of global affairs at Google and Alphabet and he believes that technical expertise sharing is going to promote responsible AI innovation Microsoft's vice chair and president Brad Smith he reinforces a company's responsibility to ensure AI technology is safe secure and remains under human control and then of course Similarly, we have Anthropics CEO, who is Dario Amodi and Anna Makenju and the Vice President of Global Affairs at OpenAI.

7:26That was Anna, I believe. But both of them emphasize the profound potential of AI to really benefit society and make oversight and governance of AI developments imperative. So in the coming months, the Frontier Model Forum is going to establish key institutional arrangements, including a character, a charter, governance structure, and funding mechanisms. In addition to this, it plans to consult with civil society and governments on the design of the forum and on meaningful ways to collaborate. And the forum welcomes the opportunity to contribute to existing government and multilateral initiatives like the G7 Hiroshima process and OECD's work on AI risk standards and social impact and the US EU Trade and Technology Council.

8:11So additionally, the forum is going to try and build on the work of existing industry, civil society, and research efforts that are kind of across its work streams. And so I think that really in an era where AI is obviously evolving super, super rapidly, I think the initiative by some of these big companies marks a pretty important step towards ensuring responsible and safe AI development. and obviously it's proving that they are at least openly putting this as top of mind. Obviously they've all been to Congress and they know that Congress is going to absolutely have a heyday if there are any missteps here so I think they're all trying to be very careful here and so I think it's going to be a very interesting thing to continue to follow as this rolls out and to see what the implications of this council are and this forum are and definitely will be interesting to see if this is a boon or if this kind of it has negative impacts on the overall ai landscape in general