Member-only story
The Bots Are Back In Town
GPT-2 Means Stronger AI
I’m wondering what advances are here now that can be employed to scale operations of bots. We’re coming up to prime time bot season and if we can look at advances in AI that could be used now that weren’t possible four years ago.
Unlike four years ago, we’re in a pandemic, economic crisis, and a reckoning with racism. If political dissent and polarization are measures of success by a foe, then Russians might be getting tired of winning. They’re about to get more tired.
The ability to scale up bot operations is much easier now than in the past. Sentiment analysis on Tweets has become so easy that you can potential look for trends among sentiment and test out which content leads to more negative sentiment. It’s possible to track engagement and target dissenting voices.
These tools can be used to track down where a human manipulator can have the most impact. New tools can go further and use natural language generation to create fake news. Enter GPT-2.
GPT-2 is part of the Open.AI movement to democratize AI. While bad actors might necessarily be using it, they could develop similar and use it to overwhelm us with things that seem real. Even a little bit of bulking up the story with quickly created supporting websites could be convincing enough for the regular Twitter user.
Experts will be able to tell which text is machine created, but most of us won’t be able to. Worse, legitimate voices that are just angry will be falsely rejected as fake.
We’re going to need new AI to help us decipher truth. This is going to be front and center over the next four months.
