I, for one, welcome our new AI overlords

Have you ever noticed that as machines get smarter and smarter, human beings are getting dumber and dumber? Sometime over the last dozen years we passed each other in transit — the robot elevator going up, the one carrying humans plummeting towards debasement. [1]

The robots have just climbed another few floors closer to the penthouse. Yesterday, OpenAI offered the world a sneak peak at its latest human replacement flagship model, named ChatGPT-4o (not to be confused with ChatGPT 4.0). This chatbot's name ends in the letter o, which stands for 'omni'. Also for 'o my f*****g god'.

While this is not the long-predicted, we-don't-need-no-stinking-humans Artificial General Intelligence Hollywood has been warning us about for decades, we can definitely see the AGI from here. 

GPT-4omfg will be all anyone who writes or talks or thinks about AI will be writing or talking or thinking about over the coming days (at least, until Apple or Google reveal something even more impressive/terrifying). And we'll all be saying the same things, because we've all seen the same demo videos. The tool isn't available for public consumption yet. But it will be, soon enough. 

O, know

So what can GPT-4o do?

It can accept input in a variety of modes – text, speech, audio, images – and respond in kind. So you can ask the mobile version to look through your phone's camera and it can tell you what it sees. 

It can read back words you've written on a page or analyze numbers in a printed chart. It can look at your facial expressions and identify what mood you're probably in. It can tell you whether that stupid hat you've got on is appropriate to wear for a job interview. (It isn't.)

It can translate between dozens of different languages in real time. [2] It can tell bedtime stories and teach your kids math. It can look inside your fridge and tell you what to make for dinner. It can look at a broken appliance and tell you how to fix it. It can be overly dramatic (Nicholas Cage, watch your back). It can sing a duet with another copy of itself. Not very well, but it can.

It can even be sarcastic. [3]

And that barely scratches the surface of what GPT-omfg is likely to accomplish.

But what's most striking to me is its ability to respond in a much more human-sounding way. (Even more so than Hume's empathic AI, which sounds like the coolest teacher you ever had in high school.) The speed of its responses, its ability to detect changes in your tone and react to them, even its speech mannerisms and random utterances, are virtually indistinguishable from carbon-based life forms. It really feels like there's a tiny person hiding inside the box — the AI equivalent of a Mechanical Turk.

Of course, these are all demos. How well GPT-omfg works in the real world is yet to be determined. Glitches are inevitable, hype is eminently deflatable. Even so, it's clearly a big step forward.

The other striking thing is that OpenAI plans to make this flagship model free to all comers. (Unlike ChatGPT 4.0, which costs $20 a month.) And that's kind of a big deal. It makes AI a thing that the 7 billion people who have smartphones can use, every single day. And that's the point.

AI is eating the world

I look at all of this, and I am reminded of that famous Twilight Zone episode, "To Serve Man." 

In case you missed that episode, it's the one where 9-foot-tall aliens with hipster goatees and light-bulb-shaped heads come down to earth and solve all of humanity's problems — they stop all wars, end hunger, and turn our planet back into the bucolic paradise it was before we screwed everything up. They also generously offer to take all earthlings on a vacation to the Kanamit home planet, while leaving behind a book written in an inscrutable code that takes human cryptographers months to decipher. [4]

The book's title is To Serve Man, and... if you don't know the twist at the end, I'm not going to ruin it for you. Suffice it to say, there's a wee bit of a catch. 

So that's what I'm wondering, What's the catch? What's in it for OpenAI? Why are they being so nice to us?

The obvious answer is that OpenAI wants the world to use its AI products instead of Google's or Meta's or Microsoft's, and the most effective way to do that is to give them away for free. 

It's also possible that OpenAI has some monetization strategies it isn't talking about yet. Remember when social media didn't come with commercials? 

But I think an underlying reason is that the company wants us to experience warm fuzzy feelings when we think about both OpenAI and artificial intelligence – more importantly, to not think about the hundreds of millions of jobs it will obliterate, the insane amounts of electricity and water required to train large language models, or who's responsible when AI is used to make life or death decisions, like where to direct a drone strike or whether it should crash your fully self-driving Tesla into an oncoming school bus or a gaggle of nuns. [5]

Obligatory caveats 

Here is where I'm obligated to say that AI also does some amazing things that are beneficial to society. No technology is evil in itself; it's how the technology is applied that determines whether it's a net positive or a negative. (And if we're lucky, it's 50/50 or better.)

For example, country star Randy Travis lost his voice in the aftermath of a stroke 11 years ago. Now, thanks to some AI magic, he's just released a new song. [6]

Programmers used vocal samples of Travis's voice taken from his 30-year recording career, took another singer's version of the song, and used AI to apply Travis's virtual voice to it. It's now number 45 on Billboard's Country Airplay Chart.

Not too shabby for a robot.

What can AI do for you? Post your requests in the comments below or email me: [email protected].

[1] Sorry/not sorry.

[2] We can now all tick "universal translator" off our list of Star Trek Inventions We Really Wish Were Real. Can transporter beams be far behind?

[3] Yes, I will soon be out of a job. Currently taking suggestions for alternative careers.

[4] Fun fact: The lead alien was played by actor Richard Kiel, famous for playing the Bond villain Jaws in both Moonraker and The Spy Who Loved Me. See, this newsletter can be educational. It's not all just dumb jokes and dystopia.

[5] Allegedly, such decisions will be governed by the principle of Humans in the Loop, though I suspect in practical terms it will look more like Jesus Take the Wheel.

[6] Hat tip to faithful reader MC for pointing me toward this.