Sam Altman, CEO of OpenAI, recently stated that he doesn’t want new AI releases to be surprising because that increases the likelihood of a world-wide freakout. Well, after yesterday’s announcement from the AI giant, people are freaking out.
During a 30-minute presentation on May 13th, OpenAI revealed their next-generation foundation model, named GPT-4o. And, like, wow. There’s a lot of surprises to unpack.
First of all, GPT-4o’s voice mode sounds like a person. It sounds like Scarlett Johansson, to be precise. And the tech bros are all over that.
Not only can the model express a wide range of emotions, it can also modulate its voice to sing, do robot impressions, and… flirt?
*deep, optimistic breath*
Second of all, the new model is very efficient at multi-modal processing, allowing it to reason with text, images, and video in real-time. Hypothetically, if it were integrated into a wearable like the Humane Ai Pin or Meta Raybans (spoiler alert: this is already happening), you could converse with the model as if it were a person standing beside you, seeing what you see, hearing what you hear… remembering everything.
*attempt a reassuring smile*
Finally, because of the new efficiencies, OpenAI is allowing everyone to access GPT-4o for free via its ChatGPT product. That means all the mass-market normies are going to experience a quantum leap from GPT-3.5-turbo to something akin to the movie Her (in which, perhaps, not-so-coincidentally, Scarlett Johansson stars as an AI voice interface). Jarvis, Cortana, Ava - whatever AGI fantasy you prefer, it’s closer than ever to becoming a reality.
So… whew. Now let’s talk about why all of this is still a good thing.
We like things that make us seem cool within our circle. This is pretty basic human psychology. For example, the iPhone’s popularity skyrocketed when the “status symbol” narrative took hold - to the point that, despite the phones costing more than some desktop computers, we will often sacrifice other important things to secure our place in the “blue bubble club”.
Up until now, there has been little hope for gadgets like the Ai Pin and Rabbit R1 to achieve true sex appeal, the former being painted as out-of-touch by popular reviewers, and the latter shipping with a design so boldly elementary that it’s hard to take seriously outside of the retro-tech San Francisco bubble.
A seamless, clever, flirtatious AI assistant could change the narrative on both counts. Imagine you’re at a party with a friend and they ask about the Chiefs game. You fire up your AI assistant to give you a summary.
Before GPT-4o: you wait 10 seconds for a reply, awkwardly shuffling your feet and looking around at nothing in particular. When the assistant begins speaking, it’s overly formal, wordy, and robotic. You cut it off because it already gave you the score and it feels more uncomfortable than helpful.
After GPT-4o: the AI doesn’t miss a beat. It responds teasingly, within milliseconds: “you call yourself a Chiefs fan? C’mon, you should know this! Chiefs won __ to __. It was pretty close until halftime, when they scored off an interception. The other team never recovered.”
Feel the difference?
You may call it subtle, but an AI that can interpret and emit social cues well is the difference between social bonding and ostracization. GPT-4o marks the beginning of the “Cool AI” paradigm. And, as any capitalist knows, mass market acceptance incentivizes more innovation - faster than ever - in the near future.
Liking this so far? Subscribe for free to receive my techno-optimistic essays each week.
More efficient healthcare
Any doctor will tell you that healthcare is more than book knowledge - it is deeply social. Patients don’t always say what they mean, but their tone and body language will often betray the truth.
GPT-4o is the first time we’ve seen a model interpret nonverbal cues and respond emotionally with low enough latency to be convincing. With a little work in humanoid robotics, we can envision a near future with AI-powered nurses doing low-risk triage. A few years after that, we might see robotic specialists doing surgery.
Believe me: as a father who just spent 4 hours in an understaffed ER with my 1-year-old at 11pm, this future cannot come soon enough.
Check out this demo, and what I’m about to say will sound more realistic.
Education is about to change forever. The decades of burnt-out, underpaid teachers preaching their own imperfect one-size-fits-all curriculum are over.
My children will grow up in a world where “school” is a one-on-one private tutoring session with world-class experts. They will learn at their own pace, benefit immediately from cutting-edge research, and ingest curricula designed specifically for their personality. They will no longer graduate from “K-12” to “university”, since everyone’s trajectory will be unique; instead, they will be trained and evaluated by AI with personalized metrics. All schooling will be free.
Of all the possibilities that emerge from the OpenAI demo, this one brings me the closest to tears. I always loved school, but I was an outlier. Most of my friends hated it - especially math - and that resentment has followed them into adulthood, reducing their inclination towards lifelong learning.
My kids will love learning, and that will profoundly influence their trajectory.
A better life for the disabled
Again, please check out this demo for a better understanding of what I’m talking about:
Be My Eyes is a company focusing on applying AI to improve the quality of life for blind people. While a personal tour guide would be expensive and unavailable to most blind people, AI can fill the same role for mere pennies.
Beyond blindness, consider all the other potential applications of this model if wrapped with need-specific technologies. Imagine GPT-4o controlling a virtual sign-language display, or creating colorful, visual representations of sounds and smells. Imagine someone who lost their voice being able to speak again - with all their original emotional range. Imagine a sound-filter for the autistic that can relay audio information from the outside world in a way that is easier for that person to understand.
This is miracle-level stuff for a significant portion of the world.
Not to re-hash a cliche, but AI will only become a force for evil if we use it evilly. While it’s easy to recall evocative Hollywood-level fantasies, these fantasies do not determine our destiny, and we shouldn’t let them burden our minds with anxiety. Those stories are, at best, a black mirror, deliberately reflecting the darkest recesses of humanity in order to stir up entertaining controversy and drive movie revenue.
The reality is that humanity, on average, is mostly good with a little bad, and that means that AI will be used mostly for good - with a little bad.
I’ll close with a melodramatic quote from Albus Dumbledore:
“Happiness can be found, even in the darkest of times, if one only remembers to turn on the light”.
Thanks for reading The Age of Abundance Newsletter! Subscribe for free to receive new posts and support my work.
