Thoughts from Ryan Barton on the rapidly progressing landscape of AI
Amid AI headlines, let’s take a moment and think about how we are thinking about AI.
If you’re like me, you tend to think of ChatGPT as a fascinating tool, and the ongoing development of AI as another disruptive technology. Something new but familiar: just the latest in a long line of disruptive technologies.
My decades as a technologist has given me a strong prediction sense of how technologies will evolve. I have an expectation of how a technology shows promise, initially disappoints, then develops, attracts early adopters, then sweeps the broader culture.
I have come to realize that I held flawed – and even dangerous – thinking! ChatGPT isn’t that. Thinking about it as “ChatGPT” masks the reality.
The truly reality-bending, mind-warping, society-shattering innovation is that beneath ChatGPT is intelligence. ChatGPT is a tool. The intelligence beneath it is what will seismically alter everything. And that intelligence will not follow the well-trodden path of technological innovation. It is something fundamentally other.
If you’re like me, you played with ChatGPT when it was running on GPT-3.5. It was cool, but the novelty wore off quickly. There is clearly some utility for all of us, but the mistakes were also obvious and a bit ridiculous. The upgrade to GPT-4 is startling. So much that it’s easy to miss how powerful this tool is already, and how enormous the growth has been in just a few months.
GPT-4 has grown so much smarter, it could be considered a fundamental different level of intelligence. And it is an early sign that the hard problems of AI are being overcome at rapid speed.
GPT-4 as it sits is enormously powerful. Artificial General Intelligence (AGI – the general problem-solving intelligence that is the stuff of science fiction) is coming. It won’t be a straight line of growth to AGI, as there are key thresholds to cross, with very hard problems to solve. But it is coming. In years, not decades.
So what? We are used to technological innovation. Every month, there is some new tool, new device, or new app that helps us get more done faster, more easily, more cheaply. So AI is now growing rapidly.
What’s the difference?
The difference is that it is not a tool that is growing. It is an intelligence. A tool accelerates one thing. Intelligence is an omni accelerator. AI represents the capacity to accelerate the solution to many problems. It is the power of thinking.
You and I must stop thinking of this as yet another technology. We must learn to frame this as a species.
You and I have adapted to new technologies. We have become skilled at it. Never before have we adapted to a species more intelligent than we are.
My friends, we are building infant superintelligences who will become giants. And they are being raised in a lab, not a family.
These child AIs aren’t that capable, but they give us a glimpse of their coming form, hurtling towards us from the future beyond our sight.
GPT-4 is something of a fundamentally different order. It is an intelligence, but not one that runs on a carbon substructure, nor one that evolved, nor one that is directly connected to reality in any meaningful way. But an intelligence, nonetheless.
The species framing helps me to hold the following in my mind:
- GPT-4 (and the many AIs that will follow) is an intelligence far more widely read than any human in history. It can already solve advanced problems in more domains than any single human being (it can simultaneously write beautiful poetry, diagnose rare medical conditions, and pass the bar exam). In many domains, this species is the smartest and best-read intelligence we have ever encountered.
- The speed of AI’s processing is much faster than you or I can think. Imagine that this species can slow down time to think for a year, while a minute passes in real time.
- The pace of growth in capability is far outstripping anything we have seen in technology to date. We have gotten used to speeds doubling every 2 years… this is closer to 10x capability in 6 months. While AI development will come in spurts (there are hard problems ahead), the pace is still difficult to comprehend. It is challenging for any of us to predict its impact, thanks to the blistering speed of change. These baby AIs are gaining intelligence minute by minute.
- Chat is just one modality. It can currently text and create images. But what’s coming soon (2024?) is generated video, speech, and personas. The communication skills of this species will soon be far beyond text.
- Innovation in other fields will merge with AI rapidly. Robotics will give it a body – lots of bodies. The flood of investments into anything AI-enabled will be prodigious. AI will become integrated into nearly every industry and facet of life. This super-intelligent, fast-growing, cloneable species will soon be offered jobs by every political and business leader on the planet.
- It is exceedingly difficult to slow this down. The race of capitalism dictates that companies must innovate or perish. The international military arms race ensures the invention of AI weapons that threaten all of humanity. It will take great societal will to keep these AIs from growing into their gigantesque form.
AI is not like any technological invention we have seen before. It is the birth of a new species.
There are many articles outlining the current technical specifications, current capabilities, and the industries it will soon touch. But none of that will help us get ready unless we can frame our thinking appropriately. And unless we can reckon with our responsibility: To properly parent these AIs, to orient them towards truth, and to integrate them into society wisely.
The coming disruption is unlike anything we have experienced before, so get ready we must.