Yesterday's AI thoughts

I posted a series of microblog posts yesterday with a common theme of trying to understand what is going on with OpenAI. This is a company with a lot of drama, nearly imploding last year with the board and CEO shakeup, and more recently alternating between amazing demos and dumb mistakes.

I got a lot of pushback about one of my posts in particular. I’m even seeing people want to leave because of it. This is disappointing to me, especially since I think I’ve gone out of my way to have a balanced approach to AI. We have a global setting to disable everything that uses AI in, for people who are against the technology on principle.

Here are the relevant posts from yesterday so you can see them more in context:

When your company becomes the enemy, all that matters to people is what feels true. OpenAI’s Sky voice shipped months ago, not last week. We hear what we want to hear. OpenAI mishandled this, no question, but most likely Her is ingrained in Sam’s head vs. intentionally ripping off Scarlett.

In the last 35 years, there have been a tiny number of truly revolutionary technologies that change everything: the web, mobile, and artificial intelligence. We can fight it, or we can guide it. But trust has eroded. To succeed we have to rebuild it. Move fast and break things will be a disaster.

Any chance that WWDC will have a live keynote this year? In the last couple weeks, we’ve had… OpenAI: live. Google I/O: live. Microsoft: live. To balance AI we need to lean in to human creativity, and a pre-recorded 2-hour advertisement will never feel as alive or engaging as a human on stage.

As someone who usually supports OpenAI, I’d still welcome an actual lawsuit from Scarlett Johansson about the voices. For one, I’m a huge fan of hers, but also I’d genuinely like to know if anything shady happened at the company. Dishonesty will cast a shadow over everything the API touches.

I also tried to clarify a few things in replies on those blog posts to other people’s points:

Sam was clearly inspired by and obsessed with Her. I don’t think it was subconscious, but that also doesn’t mean they sampled her voice explicitly.

The board firing Sam Altman is looking more and more rational. Which is why I think this should be an “all hands on deck” moment for OpenAI.

You’re right on the “total” lie, I shouldn’t have phrased it that way, because anything misleading in the OpenAI post would be dishonest. I was trying to respond to folks who are saying that when Scarlett declined to lend her voice, OpenAI copied it anyway. I don’t think that’s true. If I’m wrong, I’ll stop using any tech from OpenAI.

They already have the technology to actually clone someone’s voice, which I assume they didn’t use here because it would be an even closer match.

Did I get it wrong? For reference, here is OpenAI’s blog post about hiring actors for the voices, and Scarlett Johansson’s letter.

Perhaps I shouldn’t have blogged about this, but it’s my personal blog where I explore a range of topics. I do not run my blog posts through a PR department, and I think most people appreciate that blogs should feel authentic and human, even when they disagree.

Manton Reece @manton