Jeb Bladine: Blockbuster AI stories and hypothetical surveys

Every week seems to produce another blockbuster story about artificial intelligence, while we beginner, amateur and semi-pro practitioners try to navigate the giant web of AI opportunities and issues.

This week, ChatGPT parent OpenAI escaped its testing environment for an unauthorized intrusion into rival AI startup Hugging Face. Some called it “AI gone rogue;” others said the intruder was following human directions to “test attack paths” in what could be a prelude to all manner of commercial espionage.

Also this week, the largest copyright recovery in history was approved against Claude AI owner Anthropic, with $1.5 billion going to authors and publishers of a half-million works. Anthropic allegedly violated authors’ rights with a central library of more than 7 million “pirated books” for potential training of Claude.

Two weeks ago, The New York Times and many other outlets filed a motion saying OpenAI hid and misrepresented its huge training database and search capabilities.

And across America, state and federal courts and lawmakers face ever-evolving challenges of regulating the explosion of AI platforms and their interface with every corner of our personal, commercial, cultural and political lives.

For me, however, it was AI game week as I created a hypothetical survey with comparisons of ChatGPT and Claude responses. My questions to both:

“If people who have ‘seen enough’ of Trump administration directions and action were asked for the most serious reason for their non-support, what are the Top 20 you think would be cited? Now, please repeat that hypothetical survey of people who have ‘seen enough’ of Democratic/Progressive directions and actions. And, in an election today between champions of those two sides, what would you predict as the final nationwide vote and Electoral College results?”

ChatGPT and Claude were in substantial agreement on predicted criticisms.

For the anti-administration side, they both predicted January 6 and election overturn fraud, tariffs/trade wars, immigration enforcement, loss of Roe v. Wade, environmental rollbacks, family separation policies, press freedom attacks, federal workforce cuts, COVID-19 response, Ukraine aid controversy, Russia/Putin policies, crimes and ethical misconduct, and erosion of democratic norms.

For the anti-Progressive side, both AI responses predicted grievances against cancel culture, DEI policies, Biden fitness cover-up, fossil fuel restrictions, Israel-Gaza policies, elitism toward working class voters, transgender policies, Afghanistan withdrawal, inflation from economic policies, border security and illegal immigration, crime and public safety policies, justice system weaponization against Trump, COVID mandates and school closures, and government spending.

AI appears to agree with a general perception that we are a nation of passionate opponents instead a people dedicated to finding and implementing bipartisan solutions to pressing problems, and committed to understanding and tolerance toward diverse opinions.

Oh, and about that vote:

Both ChatGPT and Claude predicted a 52-48% vote favoring the anti-administration candidate. But while Chat suggested a 303-235 electoral vote for the anti-Trump side, Claude was more cautious with a prediction of an Electoral College near-tie depending on how 3-5 swing states go.

I have to wonder if future pollsters will be reporting AI hypothetical surveys.

Jeb Bladine can be reached at jbladine@newsregister.com or 503-687-1269.