A Watershed Time

The general public is not aware of what is happening.

A recent paper in Science explained how researchers used machine learning to create immune system “T cells” to kill cancer cells, and – crucially – keep on killing them.

Another paper chronicled how machine learning was used to design proteins from scratch – designed with requirements to bind to specific biological molecules in order to initiate immunological and other action. We are entering the time of custom protein design – something that was impossible ten years ago, before machine learning was shown to be able to accurately predict protein folding.

Biological research is transitioning from being a lab activity to a computational one. That has the potential to massively accelerate it.

Another game-changing method is the ability to create “labs on a chip” – so-called “microfluidics” systems – and mini-sized biological ecosystems, so-called “assembloids”. These can often replace the use of animals for very early stage research to test ideas, iteratively. Importantly, they can be manufactured at scale and disposed of, without any of the ethical and logistical issues that come with the use of animals – and this can be done extremely rapidly. We are seeing the arrival of robotic labs – so-called “cloud labs” (an example is Emerald Cloud Lab) – that researchers program remotely, and all the lab work is done by machines. Researchers can focus entirely on thinking rather than doing, and run experiments again and again simply by typing some code.

Biological research is transitioning from being a lab activity to a computational one. That has the potential to massively accelerate it.

Machine learning is revolutionizing other areas. If you read the news, you have probably read about ChatGPT. This is the latest iteration of a class of machine learning systems that use extremely large “neural networks” – networks containing billions of simulated neural nodes. These systems literally read the Internet and learn, and are then able to have intelligent conversations and do tasks that require intelligence.

Whether they are sentient or not is not the point. The point is that they have now crossed over into the realm of being problem solvers.

For real.

This is not smoke and mirrors. It also tells us something about intelligence, because as lifelike as these systems are, their creators explain that these systems are not thinking. They do not actually understand anything. Yet they have shown that they can be extremely creative. They find connections that even humans cannot. They derive new ideas in much the same way that humans do: by applying known patterns to other things.

These systems do not – yet – have their own opinions on things. They are merely synthesizing the opinions on the Internet and applying those, and applying them really, really well, better than most humans. But they do not form their own ideas and build on those – as I said, yet. But what they can do is have long-running dialogs in which you can pose an extremely complex problem to them, and they will solve that problem – well – and explain the solution to you, in a manner that reads just as if a person had written it. And you can ask questions about the solution and it will respond intelligently.

Change Is Coming

One thing is clear. The future is not like the past. Even the near future. Just like humans never reached the Moon until they did, ChatGPT has shown, after decades of attempts at true intelligence, that machine intelligence is possible, at least in a virtual or practical sense. Not necessarily sentience, but intelligence that can create, problem solve, communicate with people, and get stuff done. It cannot replace all people yet, but there is a lot that it can do.

Now that the Rubicon has been crossed, investment will follow. Just as when SpaceX showed that a startup can create a rocket that can get to orbit, and a flood of space companies ensued, the same will happen with machine intelligence. Millions of people will think of uses for this new capability. Next year will see the rise of tools that use true machine intelligence – a first in human history, and there is no going back.

What this means is that, once again, agility matters. Markets are about to become massively disrupted. Entire new classes of products and services are going to emerge. Biotech and literally all white collar tasks are fertile areas for disruption, and there are others. We cannot even begin to imagine the ways that the advances we have described above might be applied.

Are you ready for change?