What you need to know

• OpenAI founder, Sam Altman says AI will reduce the cost of human intelligence by a factor of 1,000,000.
• Altman's role with OpenAI remains in doubt. He was fired by the board on Friday - allegedly for lying to them, though the official statement was couched in corporate-speak. 
• Over the weekend, shareholders of the business that commercialises technology from the NFP OpenAI began pushing to have Altman reinstated. 
• Prior to the drama last week, Altman, speaking in a pre-recorded video interview with Jeff Lawson, the CEO of Twilio, at Twilio's Signal conference in Singapore, said AI will be ubiquitous.
• Don't invest in trying to build incremental tweaks on top of the model, said Altman, bet on the models getting much better and invest your bucks accordingly.
• Lawson meanwhile engaged in some hyperbole of his own (though orders of magnitude less than Altman's predictions) saying his Twilio's Customer AI would help companies do 10 times more for their customers with 10 times less.
• However, when Mi3 challenged these numbers with company execs at a subsequent news conference, it was clear many parts of the brave new AI landscape painting aren't yet coloured in.
• Likewise is Twilio's claim it would only partner with Responsible AI companies. Execs were unable to identify criteria by which they would measure and assess a company as a potential responsible partner. Sam Altman, fired by his own board for withholding information from them, is one of those partners - Hey ChatGPT, define irony.

Human intelligence might lose all value in a world of generative AI, according to the former (and potentially future) CEO of OpenAI, Sam Altman, who was fired over the weekend for lying to his board.

A statement by the not-for-profit OpenAI board, couched in corporate cookie-cutter speak noted, "Mr. Altman's departure follows a deliberative review process by the board, which concluded that he was not consistently candid in his communications with the board, hindering its ability to exercise its responsibilities. The board no longer has confidence in his ability to continue leading OpenAI."

Yet media reports late Sunday suggested Altman might return after a push by shareholders of the business commercialising OpenAI's work to have him reinstated.

While details of the extraordinary corporate event remain as opaque as the algorithms that fuel much of the commercial AI in the world, the most plausible explanation - and the one gaining the most traction - is Altman's alignment with Silicon Valley's grow fast, break stuff, burn bridges approach was inconsistent with the OpenAI Board's mission to create safe artificial general intelligence.

Altman's schtick and showmanship were on full display at the Twilio Signal event in Singapore last week, where he appeared in a pre-recorded interview with Twilio CEO Jeff Lawson, during a keynote address.

"My whole view of the future is if we can drive the cost of intelligence and the cost of energy down to zero or close to zero, we create massive abundance in those two areas," Altman said. "Those are the things that limit us from everything else we want to do. Right now, intelligence is just too expensive to do that for every customer, because it's actually human. If the price of intelligence falls by a factor of a million, then you can do stuff like that."

Ubiquitous AI is basically the dial tone of business in Altman's new world. Lawson agreed Large Language Models (LLMs) would allow brands to effectively deploy virtual employees for every flesh and blood customer.

According to Altman, "Right now you hear people talking about AI companies that won't last for long. That's ok, because you can create companies now you could never create before... And there are a whole bunch of things that just didn't make sense before AI. But you have to assume everyone's going to have it and go build a business with all the inherent characteristics of a good business."

Business AI 1.0

It's for this reason Altman believed a lot of current development efforts for AI are being misdirected.

"We see a lot of companies building things that are just a small improvement on top of the model. Say the model doesn't do something quite right," he continued. "And people put a lot of effort into a very specific prompt, or layers on top of it, to get it to do something. We recommend against doing that because the next turn of the model crank just gets smarter and smarter. Trying to augment the fundamental intelligence of the model is, I think, a lot of work for low reward."

Instead, Altman urged companies to bet on the models improving. "It's very interesting to think about things that are a couple of years away from a capability perspective. What is GPT 5 or 6 likely to be able to do? What will you be able to deal with that? How can you start thinking about that now? Even if you have to have humans in the loop, or a whole bunch of other things, which I think makes a lot of sense, given current limitations that's ok if you believe that AI is going to keep improving.

"One of the best pieces of advice at YC [Y Combinator- from where Altman also departed in slightly cloudy circumstances] is to do things that don't scale. It's ok to do things that don't scale now if you assume a smarter model will fix the problem."

Altman also noted just because AI is loose in the world, that doesn't mean you get out of any of the hard work of building a business.

"It's just a technology layer. You've still got to think about how you're going to build a network effect or some sort of competitive advantage, you still got to have a great product in relationship with customers, it's still got to get better as it gets bigger, as the network effect gets stronger or the compounding advantage get stronger. But as long as you're thinking about that, and viewing AI as an enabler, you can bet it's just going to keep getting smarter."

A piece of the action

Twilio's CEO brought his own hyperbole to the stage, albeit toned down by several orders of magnitude compared to his guest. Lawson said Twilio's Customer AI-built open large language models like ChatGPT would help brands do 10 times as much work at one-tenth of the cost.

In addition to work in AI, the company has strong positions in communications-as-a-platform (CPAAS), and also the data orchestration market with its Segment CDP - the market leader by volume in Australia, as we reported earlier this year.

"By bringing together the leading communications platform and the leading data platform, we can deliver this customer AI roadmap to you to unlock this value, to make it so you're able to connect the dots for those customers using AI," Lawson said. "I believe companies who use this technology are going to get ten times better at serving their customers. And they're going to do it for 1/10 of the cost."

This, Lawon said, will be the defining factor of the next generation of companies. However, when MI3 pressed company executives in a subsequent meeting to justify the numbers, given the IT industry's long history of promising orders of magnitude but delivering increments, details were thin on the ground. It was likewise the case when asking execs to describe the criteria by which they would assess a potential AI partner as 'responsible', despite that being a key part of their AI pitch at the event.

*The author travelled to Signal23 in Singapore as a guest of Twilio.