AGI or not, we can't close our eyes
Monday, December 30, 2024
There’s been a conversation about when AGI will arrive for a long time. That time horizon has shrunk since chatGPT came on the scene, obviously. However, for a while I thought maybe these models would asymptote. I’m not convinced of that anymore. I think AGI, or something close enough, is hurdling towards us. We need to start grappling with these questions (and by “we” I mean the general public, as I’m sure many in the industry have these conversations frequently):
- How do we handle advertising popping up in AI responses? (I use search engines far less than I used to. With eyeballs spending more time with AI chatbots I’ve no doubt ads will be in responses soon - especially to keep the tools free.)
- How do we prevent the government from capturing the only companies that can afford to build competitive models? (The government is often captured by corporations and special interests so this would mean the models are one degree of separation away from also being captured by these interests.)
- How do we make sure those companies don’t drive policy to ensure they’re the only players?
- How do we ensure that ideologies don’t seep into AI responses, thereby nudging the population in directions that the powerful deem “correct” or “good for us”?
Essentially there will be (is?) a power struggle for influence how these models respond to all types of queries. As I see it we basically have one shot at preventing this level of capture – once the government and the companies are sufficiently intertwined it won’t be undone.
We have to get past the fear of AI. The AI apocalypse might happen – no one knows. But even if technological progress stopped today, all of the questions I laid out above will be answered in the next few years. I don’t know the answers, but if we don’t participate in the conversation then they will be answered for us and almost certainly not in our interests.