Reading The Tea Leaves On AI

Growth strategy busines trend concept.

getty

As we continue to ask “what’s happening with AI?” we also continue to ask the harder question: “what’s about to happen?”

It’s easier to look around right now, and see what we can do with the technology that we already have. Many would say that in some key senses, we’re already close to artificial general intelligence (AGI) or the “singularity,” a somewhat dreaded threshold where AI advances past us as humans, cognitively. But it’s harder to extrapolate into even the near future, for instance, 2027.

What do we have in the way of this new technology scrying?

Well, I want to go back to something that I’ve covered before, one of the more colorful, elaborate sets of predictions from Daniel Kokotajlo, who has appeared at some of our MIT events, and a team of experts, titled simply “AI 2027”.

In this detailed scenario, by 2027, a fictional company called OpenBrain has released something called “Agent 2” and the authors present this potential outcome for the year 2027:

“With new capabilities come new dangers. The safety team finds that if Agent-2 somehow escaped from the company and wanted to ‘survive’ and ‘replicate’ autonomously, it might be able to do so. That is, it could autonomously develop and execute plans to hack into AI servers, install copies of itself, evade detection, and use that secure base to pursue whatever other goals it might have. Results only show that the model has the capability to do these tasks, not whether it would ‘want’ to do this. Still, it’s unsettling even to know this is possible.”

As I said, this set of predictions, in the form of a multi-page essay, is very detailed. Some other forecasts are broader. I’m reminded of an interview I did with Andy McAfee at a recent TED talk event, where we discussed his theories on how AI will turn out by next year.

Three Scenarios

First, McAfee made three predictions.

The first one was, in general, a prediction of prosperity, that AI will not “take away jobs,” but complement human value.

“I'm not worried about the AI job-pocalypse,” he said. “I am pretty sure that AI is the best skill delivery tool that we humans have ever come up with.”

He explained:

“I am suddenly pretty competent at writing code. Everybody has this matching tool to help them realize their ambitions, so skills entrepreneurship is going to go up, prices are going to go down…”

Another of McAfee’s predictions was “higher levels of human flourishing”.

“Our lives are being bettered because of AI, and in general, this wave of technological progress that we continue to under-appreciate,” he said.

How about the third one? In a bit of a pivot, McAfee’s third prediction was that as we see these advances, we’re going to keep whining about it.

“We are still going to emphasize the negative,” he said. “We're going to sit here, we're going to complain about the terrible bad things that are happening, and we're going to be professional pessimists and downers about this period of progress they're living through. And that, that, that negative cheerleading is unfortunately going to be led by people from fancy universities, because the way you signal how smart you are for a fancy university is by being more pessimistic.”

He also went into some of the more gloomy projections about SkyNet or hostile, lethal AI:

“We spend too much time problematizing instead of analyzing,” McAfee contended. “We pick on the thing that might go wrong, and we spend all of our time talking about this. The fact that we still have this somewhat serious discussion is, is the AI going to kill us all? This is bizarre. If you’re going to make an extraordinary claim like that, the onus is on you to provide some evidence about AI rising up inadvertently, deliberately, and killing us all. I haven't seen that evidence anywhere, but yet we're still bogged down in this nonsense conversation.”

From California to New England and Beyond

Later in the discussion, I asked McAfee about how big companies, traditional companies that move slowly, will do now that three people can do what a workforce of, say, a thousand, did before AI.

Pinpointing the source of much of the current innovation in a “small peninsula” in California, McAfee said he hopes to be wrong about the challenges that latecomers face.

“The incumbent economy, the legacy economy, the 20th century economy … they're growing, they're becoming more productive,” he said. “Are they ready for the wave of disruption that is coming at them? We're in the early innings of a whole lot of change in the business world. I'll say it one more time. We're going to be the beneficiaries of that change, but there is going to be some turbulence and some headwinds involved for historically successful companies.”

Movers and Shakers

In response to one of my questions on adoption, McAfee had this to say about a vanguard of people who find themselves at or near the forefront of today’s AI revolution:

“There's this crop of people who we probably would not have predicted in advance, who are super-enthusiastic about this new toolkit, and they're the power users or the evangelists of the people who are on the fence, or who find it super-intimidating. You know, the discussion about agents and orchestration and coordination and your markup files, all of that's intimidating, right? I still find it a little bit intimidating, largely because it changes so quickly.”

His closing remarks has to do with valuing people, giving them platforms, and eliminating barriers to AI adoption.

I think those are good goals, especially in light of what we already see. As McAfee noted, people are intimidated. They are threatened. And when they feel threatened, people don’t make the best choices. They hesitate. They lack the confidence that is going to lead us to better outcomes.

So let’s move forward together, picking each other up when we stumble. AI doesn’t have to be this travesty. As McAfee points out, the building blocks are there. It’s the builders we need.