I don’t know if hydrogen is the answer, but I have become convinced there are fatal flaws in the current path we are on. Hydrogen is the best alternative I’ve found to date.

it already is in important ways. It’s lighter, for example. And you could in theory scale it way beyond how much you could physical batteries. But attacking hydrogen does not solve the fatal flaws in magical EV+batt thinking.

Not making any predictions, but the major problems I’ve listed have gone unanswered, because there are no answers. Either battery tech dramatically improves or current thinking is very very wrong. Plan seems to be “cross fingers and hope for bette batteries”. Not a good plan.

Genuinely interested in hearing your answers. Without 10-100x battery improvements I don't see how we get to net zero, therefore a good chance hydrogen plays a major part contrary to widely held beliefs. Efficiency calculations are often wrong - hydrogen maybe half as efficient. Not a showstopper.

How do you solve slow charging times for electric vehicles? How can people charge a vehicle without charging point access at their house (it's often physically impossible in the UK)? How do existing electricity networks cope? How do you deal with pitiful EV range during the winter months?

How do you intend to store the vast amounts of energy required to rely on renewables? The sun doesn't shine at night or much at all in winter and the wind is very unreliable, even in the UK. Huge amounts of energy must be stored for long periods. How do you solve planes, trucks, or long drives?

I think recent models hold huge potential but very few companies are building great products from them. Product and UX absolutely key as AI opens up new paradigms. Copying and pasting into ChatGPT as a workflow or wedging in “summarise” or “rewrite” features into legacy apps is not the way to go.

Think we can also conclude that software engineers being 10x or even 2x more productive due to AI is unrealistic.

I use LLMs all the time in coding and I'd say I get a speed-up of about 10%, more if it's very repetitive code.

From my past research one thing I would say is that I would expect the brain’s learning methods to be quite robust and transferable. And the best way to achieve that is simplicity and scale. Recent DL does hold promise in that sense. But that does imply deeper similarity.

Fortunately, Papermill’s success is not dependent on AI improving. It sure would help make the product better, but we’re providing a lot of value to customers regardless. So I can afford to be fairly neutral on the topic.

Some folks (notably academics) are nonplussed by ChatGPT. They are very much in “stochastic parrot” territory. My issue with that argument is I’m not sure humans are too different from stochastic parrots. We just don’t know AFAICT.

"...what we know for certain is that at some point in the early twenty-first century all of mankind was united in celebration. We marvelled at our own magnificence as we gave birth to AI - A singular consciousness that spawned an entire race of machines. We don’t know who struck first, us or them."