Former OpenAI Board Member Helen Toner on the Uncertain Future of AI Regulation | StrictlyVC DC

Rebecca is going to tell you a little Bit more about Helen who thank you Helen So much for coming um and I'm really Excited about this next conversation Thank you Both yeah as Connie mentioned we're Going to be chatting with Helen toner Who many of you may know as one of the Former board members of open AI but she Does much more than that um she's also The director and of strategy and Foundational research grants at the Center for security and emerging Technology I got that all right Fantastic C set is easier C set for Short Perfect well we've got a lot to talk About today similar topics what we've Already been kind of chatting with with Other guests AI reg a AI regulations say That five times fast and policy but I Think one question to start that I've Been wanting to ask you in you in Particular for no specific reason but Can a company regulate itself when it Comes to AI I think Um I mean it's a bit of a setup isn't it I I I I mean I wrote a um WR in The Economist a couple weeks ago that I Think the answer is essentially no and I Think especially for um AI companies That are aiming to build extremely Advanced AI technology so not just you

Know particular app for a particular Sector but um companies that are you Know aiming to build artificial general Intelligence as they call it um or Otherwise really seeking to build you Know highly general purpose highly Powerful Technologies um I think that is A um uh you know a really ambitious Endeavor and one that deserves a lot of Scrutiny and one that it deserves a lot Of participation and engagement from a Broad range of stakeholders um and I Think that is an idea that is um Becoming more and more intuitive and and Widespread and I'm I'm glad to see That and just before we dive in further Taking a step back because I know Reading the headlines and sort of Reading what comp different companies Are doing and what different companies Are getting slaps on the wrist for what Other companies are seemingly getting Away with getting sued with it's made me Think time and time again is there any Regulation on AI currently like what Does regulation look like if any is There any policy this stuff falls under Or is there kind of we're starting fresh No absolutely I mean there's a lot of Regulation that that already applies to AI including you know some that that Sher Conan just just spoke about um and Often it's regulation that you know Applies much more broadly and if you

Build something with AI then it still Applies you know there's no I think Actually the FTC has said you know There's no AI exception to existing laws Um so you know in addition to the kinds Of false advertising or deceptive um Practices that the FTC would look at you Also have all kinds of sector-based Regulation so if you're building AI for Healthare for transportation for finance Um we have existing sectoral Regulators That um have rules rules for what's okay And what's not okay in those spaces and In many cases AI systems will be covered By those existing rules um sometimes It's not a perfect fit so you can get Into kind of the the like policy Nitty-gritty of well what investigative Powers do they have or well which actor Can be penalized for which kinds of Behaviors but by and large I would say There's there's plenty of Regulation um Primarily on a sector by sector basis That that certainly already applies to AI That's so interesting to hear because You obviously know this a lot of people In the room know this when you start Talking about regulation when it comes To AI it's like you brought up the Boogeyman it's like you brought up Voldemort it's like Founders don't talk About it VCS don't want to talk about it Companies don't want to talk about it And yet there's regulation that a lot of

These companies already have to deal With or sort of guidelines they have to Follow anyway so why are people so Afraid of potential regulation and why Is some of that fear potentially a Little misguided I mean I think I think Some of the fear is is not misguided I Think there are plenty of places where You can see regulations really going a Or being too heavy-handed I think you Know for example there's a whole um Whole thing in the discourse right now Around um uh at least on my Twitter feed I don't know what your all Twitter feeds Look like um around uh you know the Green transition and green energy and The ways that some of those potential Projects we might want to invest in are Being held up by environmental reviews And so you know you're obliged to go Through this multi-year sometimes multi- Deade process to do an environmental Review and if you're doing that for a Giant solar farm and the Sol Farm gets Delayed or canceled because of you know A particular species of lizard I don't Know um maybe that wasn't actually kind Of a net win for society um so I think It makes sense that people kind of have Some trepidation and especially um some Of the kind of big picture or like uh The short versions of policy ideas that Sometimes get thrown around for AI can Sound kind of heavy-handed and and and a

Lot um but I think you know with a lot Of things in policy it really depends on The details it really depends on the Specifics and um and I think sometimes The Sort of uh uh yeah the headline level Conversation about about AI policy can Turn into a big thing of well there's Regulation versus Innovation and you Have to do like not too much and not too Little um as opposed to looking at Actually there's all kinds of different Ways that um that government can be Involved also that you can have other Forms of you know governance soft Governance not just hard law um whether That be uh industry arrangements or kind Of um mechanisms uh you could have Between different trade organizations or Industry standards um there's really a Lot of options really big option space Here and so I think um while on the one Hand I am sympathetic to not wanting Poorly designed or heavy-handed Regulation at the same time yeah I do Think this sort of immediate like jump Scare at any involvement by government Of any kind uh is not not really Appropriate and what you just talked About there are all the different ways There could be regulation that isn't Super heavy-handed reminding me of Something we talked about last week when We were chatting about this in

Preparation for tonight is that you were Saying that regulation is a spectrum It's not not just it's one end or the Other and there's just so much in Between of where it can end up falling In the future and how can AI companies Today or sort of people in the AI space Work with the government and work with These organizations to ensure that the Regulations that do come in the future Kind of work for both sides like how can AI companies get directly involved as Opposed to say waiting for stuff to come Out and just saying if they like it or Not yeah I mean I think seconding what Steve was saying at the beginning of the Evening I think it's valuable to be um Involved in conversations to be talking With folks um on the hill also in the Executive branch um you know I think Most of the room is based in DC so I'm Sure that that this is something that's Familiar for you um the policy World Works in large part on the basis of Relationships and trust and uh you know There's a lot of judgment calls that Need to be made and so if you can be um In some of those rooms and talking to People about what are the technical Realities of the work that you do or the Work that companies you fund do um I Think making sure that there that tight Coupling between you know what is um What is some Bill in Congress trying to

Achieve and how what is it in practice Going to do are there going to be Unintended effects because they didn't Realize that some definition actually Brings in a bunch of stuff they didn't Mean or you know other things like that Um so I think kind of thinking of it as A a proactive productive collaboration Between different people with different Kinds of expertise um and bringing the Expertise that you have is really Valuable um I think thinking of it as uh Either something to kind of ignore Or as sort of an an adversarial fight Where you have to be kind of uh trying To get you know the other side to do as Little as possible I think we'll be less Productive um I also think it's worth Really thinking about uh uh you know the Different ways that this could play out Um especially if you know right now We're coming off the back of a few years Maybe you know a decade or so of AI Continually getting more sophisticated More powerful more capable more widely Used um and think in a lot of ways to me The default is Congress right now I Don't know if anyone's noticed not super Functional um not super good at passing Laws um unless there's a massive crisis And I think one thing that I sometimes Um wonder about is whether people who Are trying to you know minimize the Chance that there's any AI regulation

Now whether they're thinking through the Fact that if in fact there's nothing Passed and nothing passed and nothing Passed and nothing passed and then Something happens with AI which at some Point it will it's going to be a big Powerful technology something will go Wrong at some point um if the only laws That we're getting are being made in a Knee-jerk way in reaction to a big Crisis is that going to be productive um Because that you know that is the Default path is you know nothing nothing Nothing until there's a giant crisis and Then a need jerk reaction and so I think Sometimes um I think some of the sort of Smarter and more thoughtful actors that I've seen in this space are trying to Say okay what are the pretty light touch Pretty Common Sense guard rails we can Put in place now to make future crises Future um big problems less likely less Severe and basically make it less likely That you end up with um the need for Some kind of uh rapid and poorly thought Through response later um which in some Ways I think again is the default given How um how bad strongest is at at Legislating right Now and I know a couple weeks ago um With open AI open AI kind of running Into that issue with Scarlett Johansson And did they use her voice did they not We're not sure of course there's a lot

Of incriminating evidence but we're not Sure we're not sure um but I when people Were talking about using that as an Example of like why we should have Clearer regulation around stuff like This so that people don't make these Mistakes I was reading a lot of comments On Twitter and these were from real People some of them even being VCS I've Interviewed in the past like not Bots And they were like well every other Sector we've just moved fast and break Things like why does it matter so much In this case that this is now wrong like Oh Facebook did that was that wrong Which now we know years later that that Was wrong but like at the time it was Like that was wrong like why why is it Wrong if AI companies move fast and Break things I mean I think I think some Amount of that is right I think you Don't want to be uh trying to predict The future too much and in in many cases Waiting until there's a problem so you Can know what problem you're trying to Solve is is very reasonable um again I Do think there's a bit of an exception For the very small number of companies That are trying to build extremely Advanced AI systems um telling you we're Going to build human level AI super Human level AI we're going to do it in The next 5 years like that's a pretty Different kind of pitch for your company

Than oh you know that book of faces that Your University gives you we're going to Like do a digital version of that that's Kind of kind of a different different Thing um so I think it's it's kind of Reasonable for the public for government For sort of society at large to look at Those companies and be like okay you Want us to trust you why should we trust You like convince us that we should Trust you and I think um the different Companies have somewhat mixed track Record over time of um of earning that Trust um and I think that the importance Of of of having that trust and earning That trust is only going to go up again If the sophistication of the technology Keeps keeps improving definitely and I Know a lot of the conversation around AI Policy and regulation has focused on Like the big names that we all know that Are already established in the space Like the microsofts the Googles and the Open AIS but of course we're in a room With VCS and startups and there are a Lot of startups trying to break into the Space either they're looking to start a Larg l build a large large language Model again say that five times fast a Large language model up from the ground Um themselves or they're looking to sort Of build technology off of say open AI Um model as well and how can they think About not just the regulation of today

But potential regulation that actually Seems realistic for the future so that If they're kind of just getting started They're making sure they're not building In a way where they could have more Problems down the line if regulation Goes into place than if they had built Say gone the other four the road how can Startups think about those kind of Things even though they haven't come Into effect yet yeah a couple thoughts Here I think one is to be very Deliberate and intentional about your Choice of sector and use case um so There's just a really I mean a huge Problem with talking about regulating AI Um is that AI can be anything you know AI can be in literally any any sector You can imagine um and so I think there Is a real difference if you're using AI For to do elder care versus using AI to You know recommend videos or build a Spotify playlist um there's a real Difference in the level of scrutiny that You should expect to be applied to your Product um and I think that's not to say That no one should build in those kind Of higher scrutiny higher regul Regulation sectors there are often some Really really important sectors but if You're going to be in a sector like that I think you you just are going to have To be um very thoughtful from the Beginning you know with things like uh

Documentation I think is really Important if you're in a a place where You think you might down the road um be Subject to any kind of Regulation Because you're in a kind of higher risk Higher impact sector um keeping track of What are you training how are you Training it how are you testing it what Are you doing with your data how are you Updating it um Which models you have Deployed in which products at which time Um you know the stories you hear from People who've been on the inside of some Of these big companies is they often Even like really big really respected Companies um is they often don't even Have good records of like oh wait hang On which combination of models is Running that particular like product Right now now um so I think those would You know be sort of two two suggestions That would come to mind for me um first Be thoughtful about the sector you're Going into and what you know level of Scrutiny you should expect there and Second um if there might be scrutiny Then think of documentation as kind of Your first line of defense and just Having um well-kept records about about What you're Doing and I know it's come up it came up In both the conversation with Steve as Well as with lenina con as also but Startups have different resources than

Microsoft we all know that so it's like Thinking of being a part of the AI Conversation how can smaller companies With less resources make sure that they Too can work into this conversation and They too can kind of map out the Potential regulation policies that Affects them as well even if they don't Have as deep of pockets to kind of Influence where it goes yeah I think it Comes back to those those relationships To being available to being interested For policy makers often from the Perspective of policy makers they know That they don't understand the Technology and they really want to be Able to talk to people who do um but It's hard to know who that would be Often the easiest people to find or you Know a friendly neighborhood lobbyist um So if you can be there uh if you can go To events if you can kind of strike up a Conversation um yeah I I don't know if There's a better way than than that I Mean and then of course there would be Other things like um uh you know joining Groups um being part of communities like This uh which let you also um kind of be Part of something larger as a collection Of of sort of smaller actors Um I think can also be Valuable and I know the election again Has come up in all of these sessions so Far but we all talked about it we all do

Know it's election year and similar to What I mentioned with Steve of course Who wins president will have an impact On things like this but almost more so It's who gets voted into the house who Gets voted into the Senate from the States maybe you're not you thinking California thinking New York maybe from The states that aren't don't come to Mind when you think of AI and so how What should we expect from this election Not necessarily in terms of who's going To win but will this election Potentially does it even have the Opportunity to potentially speed up Regulation do you think it's going to be More of the same like how do you think This year could end up impacting say the Conversations we're having around this Next year yeah so I mean I I tend to Think about this a lot from the Perspective of the space that I mostly Work in which is more the National Security Foreign Affairs space and um I Think one one big effect uh one one Effect that people noticed from the last You know eight years which I think is Going to be relevant for the next term Whichever kind of President it turns out To be is um that you know a potential Trump presidency is just going to be Much more unpredictable much more Erratic um than a Biden presidency in Many ways the sort of Mainline policies

May not be that different necessarily Depending on the issue um but I think The sort of process and regularity and Predictability of the Biden years has Just been much greater than what we saw Under Trump and so I think uh sort of Expect the unexpected did um if we do See a second Trump term and then a Second big thing that that um I'm Keeping an eye on is yeah as you Mentioned not just the presidential Level but kind of uh in Congress more Broadly how I mean essentially the way I Think about it is is how functional is Congress how able to do its job is it um I think you know a friend of mine who Spent many years on the hill uh told me That he thinks um the house is currently More dysfunctional than it's been since The Civil War um so that's like you know Um A big claim um so you know does that Worsen in the next Congress does that Potentially improve a little bit um There's been this whole big process run By um uh leader Schumer in the Senate um Around you know running these Insight Forums about a I'm sure many of you Followed them um writing up a whole big Report about the findings of that um is There uh does that kind of get carried Over does that uh continue does that Sort of is that an initial setup for Congress to step in and build on that

And make laws and uh and kind of carry Things forward or does that sort of that Folder get dropped on the ground and um Lost in the shuffle and and we're Starting from scratch um and I think That will be um partly result of which Parties are in control but also a result Of um just the kinds of U members and Senators that we get um uh you know do We have as chaotic a speaker election um And speaker you know continuation as We've had this time round um and of Course you know what other priorities do They choose to to pick up so those are I Think that kind of the two big things That I'm that I'm Watching and thinking about sort of past Areas that we have been slow to regulate On things like social media things like Crypto I know you talked to crypto Founders today they're still not 100% Sure exactly where regulation is going Regarding their field either do you Think AI will be different like will History repeat itself will we the the um Current government or the future Government sort of wait too long or as You mentioned earlier wait until Something happens to really decide to Take a stand or really decide to fix the Issue or do you think we've learned from That do you think we'll see regulation Maybe not soon but maybe a little Earlier than we have in the past like is

Now any different than some of the past Like Tech crazes that have come in that We've then kind of tried to figure out As they're Happening to be honest I think everyone Says that they've learned and then I They don't seem to act very differently So um I if I had to bet I would bet that we'll Probably see something similar and also Similar in as much as you know Europe Has passed major AI legislation um I'm Sure plenty of people in the room are Thinking about the European market and What compliance with the AI Act is going To look like not to mention the digital Market act um the Digital Services act Um and so it may well be that kind of The where things go with AI mirrors Where they've gone with something like Data privacy where there's actually Quite different rules in place in Different jurisdictions um Uh you know in in consumer privacy we Saw um California eventually passing Pretty significant legislation so There's also question would that Dynamic Repeat itself here um I think it's it's I think there's a lot of ways it could Go I do think will depend what happens With the technology if we do see um sort Of things continuing to um ramp up and Accelerate and kind of more uh more Concerned about increasingly powerful AI

Systems versus if you know as some other Experts believe things sort of plateau And there's not that much new exciting Stuff happening over the next five years Um I think that will affect where things Go um but yeah if I if I if I had to put Money on it I would probably put money On um there not being enormous new Changes at the federal level um and then It's just a question of what else uh What else might happen around the Edges and um Bas we just have time for One last question and I'm just curious With everything that we've talked about So far and sort of all the experiences Had working with both open Ai and some Of the other companies you work with Through the work that you're doing now Are you optimistic we're going to get it Right both on the regular side and on The business side My overall picture With AI policy and my answer to this Question as well is just like huge Uncertainty and I think anyone who is Sure one way or another sure that things Are going to go fast or that they're Going to be really slow sure that They're going to be amazing or terrible Um I I think it's misguided to have a Ton of confidence either way so I think Um I think there's a really wide range Of potential Futures and um in one way It's going to be very exciting to get to See where things go um hopefully it's

Just exciting and not you know not scary Or bad um but I think I think we should Take seriously the possibility that There's just a very wide range of ways That that the future could play out from Here definitely no join me in thanking Helen