The development of the lithium-ion battery is an example of how scientific discovery does not have a straight-line trajectory.
How We Got the Lithium-ion Battery
"One notable thing about the evolution of the lithium-ion battery is how hard it is to predict the trajectory of research, and how important it is to allow researchers the flexibility to pursue what they feel is promising. Whittingham stumbled across an intercalation-based battery when researching fast-ion transport through a solid electrolyte, an entirely different phenomenon. And his invention of the first lithium-ion battery cathodes was the result of a serendipitous discovery during work on superconductors. Thackeray discovered the manganese oxide cathode at Oxford ... for a year’s sabbatical so he could pursue the battery ideas he found promising. Early research on a graphite-based anode, performed by Rachid Yazami, was originally aimed at discovering a graphite-based cathode, not an anode, and Akira Yoshino’s battery efforts at Asahi Chemical were pursued in spite of the fact that company thought very little of the battery market, and only bore fruit because the company didn’t actively try and stop him. Likewise, the discovery of ethylene carbonate as an electrolyte that would allow graphite to be used as an anode was an accidental discovery by Moli Energy.
This sort of trajectory, of course, makes it hard to capture the value of research, or to have anything like a reliable, predictable path by which scientific research gets turned into marketable products. Exxon’s efforts to develop a practical rechargeable battery ultimately failed, though its research would spawn a successful battery in the fullness of time."
I don't think LLMs are suitable for chess. Witness the result of a chess game this year between ChatGPT (OpenAI) and Gemini (Google AI). The LLMs had a very poor record at picking legal moves, let alone good moves:
"Gemini's mistakes and ChatGPT's misses tell a lot of the story. One AI kept giving the other one opportunities, the other kept refusing those gifts. The good news for ChatGPT is that it made more "Good" or better moves than Gemini made mistakes and blunders. The bad news for Gemini is, well, almost everything that happened. ...
The final illegal move tally was Gemini 32, ChatGPT 6. That makes sense; it would have been crazy if the AI good enough to win was also bad enough to make more illegal moves. But it also means Gemini only went 50% in picking a legal move, while ChatGPT was over 80%."
Okay, here goes nothing. My small take on the Excel story.
The year was 1986, pre-spreadsheets. Writing up an undergrad physics experiment in 1986, I needed to do a hundred or so similar calculations and present the results in a table. Luckily, a computer science acquaintance wrote a small program to do this task for me. Thanks, Dan.
In 1989, before Excel, there was Lotus 1-2-3. Loved that spreadsheet software. My PhD task involved plotting lots of data points, including smoothing some of them. Doable with Lotus 1-2-3, probably not doable otherwise.
More recent times.
Spaghetti Excel. An engineering acquaintance told me his student summer job, at an aluminum refinery, was to check and simplify their Excel spreadsheets. Apparently they had numerous spreadsheets linked to each other. I assume the main purpose was inventory control. I know I didn't envy him his task.
"Please, not just Excel." I took a class of high school students to a uni chem lab and had a small argument with a chemistry tutor who insisted the students use Excel to plot their data. I wanted them to first do it by hand with graph paper. This would have given them a much better feel for their data.
"Rinse-and-repeat Excel". I was tutoring a construction guy trying to learn maths. His job as an assistant on a high-rise construction job involved putting lots of numbers into an Excel spreadsheet. The check for this was to repeat the process and see if he got an identical result. I thank G*d his boss made him do this.
And that's it. Helped other people to use Excel, but I'm thankful that I haven't had to spend my life inputting data into spreadsheets.
There were millions of spreadsheet users by 1986, as VisiCalc was released in 1979[0] and similar programs like SuperCalc[1] were also in use. They were both ported to IBM PC and saw significant use in the corporate world prior to and including 1986.
Wow, this guy has been writing the code for card games for 10+ years. And the number of games has certainly grown. I remember when he had only coded about 3 games. No bridge, of course; it is the most complex of card games. The main differences being that bridge uses the whole pack, which create many bidding levels, and that one hand becomes exposed after the bidding, giving greater accuracy to the play.
When I used NMR scientific instruments that cost hundreds of thousands of dollars in the 1990s, the Japanese (Jeol) machines had a far more straightforward interface than the German (Bruker) machines.
Of course, the study of how living cells function is "hard". But that doesn't mean it has to be learnt without joy. We tend to explore things we enjoy. A lot of the writer's essays [1] are about finding some aspect of a topic intriguing and following that rabbit hole.
My own research centered on one subset of functions within E. coli. I was lucky that I found a carefully engineered subset of plasmids and adaptions of E.coli, that could be mathematically modelled [2] [3]. I didn't have to know the whole functioning of E. coli. I didn't have to use mathematics beyond algebra. That is, no calculus was needed. The key task was to put together the quantitative research of about half a dozen labs. Okay, I had a "mountain" of articles to read. And it took 5 years of effort. But it was only doable, because I was modelling a carefully constrained subset of cellular functions.
A suggestion: I enjoyed one short online computer course I studied, because we were given other students' submissions to assess. I recall using both my own code and the code of another student to suggest improvements in a third student's code. I am talking mainly about small tasks such as coding a function.
To me this is just formalising what students do anyway: namely, help each other understand and complete tasks in a course. The difference is that the instructor is actively giving students access to other students' code. I found the process motivated me to get stuck into a task, rather than leaving it to the last minute.
Reading the transcript of Bill Gates's (Nov 16) interview with Yejin Choi (University of Washington / Allen Institute for Artificial Intelligence), one gets the impression that AI is still in its teething stages:
YEJIN CHOI: Usually, the smaller models cannot win over ChatGPT in all dimensions, but if you have a target task, like a math tutoring, I do believe that definitely, not only you can close the gap with larger models, you can actually surpass the larger model’s capability by specializing on it. This is totally doable, and I believe in it.
BILL GATES: Certainly for something like drug discovery, knowing English isn’t necessary. It’s kind of weird, these models are so big that very few people get to probe them or change them in some way. And yet, in the world of Computer Science, the majority of everything that was ever invented was invented in universities. To not have this in a form that people can play around with, and take a hundred different approaches to play around with, we have to find some way to fix that, to let universities be pushing these things, and looking inside these things.
YEJIN CHOI: I couldn’t agree with you more. It cannot be very healthy to see this concentration of powers so that the major AI is only held by a few tech companies, and nobody knows what’s going on under the hood. That’s just not healthy. Especially when it is extremely likely that there is a moderate size solution that is open, that people can investigate and better understand and even better control, actually. Because if you open it, it’s so much easier for you to adapt it into your custom use cases, compared to the current way of using GPT-4, which all that you can do is sort of a prompt engineering, and then hope that it understood what you meant.
How We Got the Lithium-ion Battery
"One notable thing about the evolution of the lithium-ion battery is how hard it is to predict the trajectory of research, and how important it is to allow researchers the flexibility to pursue what they feel is promising. Whittingham stumbled across an intercalation-based battery when researching fast-ion transport through a solid electrolyte, an entirely different phenomenon. And his invention of the first lithium-ion battery cathodes was the result of a serendipitous discovery during work on superconductors. Thackeray discovered the manganese oxide cathode at Oxford ... for a year’s sabbatical so he could pursue the battery ideas he found promising. Early research on a graphite-based anode, performed by Rachid Yazami, was originally aimed at discovering a graphite-based cathode, not an anode, and Akira Yoshino’s battery efforts at Asahi Chemical were pursued in spite of the fact that company thought very little of the battery market, and only bore fruit because the company didn’t actively try and stop him. Likewise, the discovery of ethylene carbonate as an electrolyte that would allow graphite to be used as an anode was an accidental discovery by Moli Energy.
This sort of trajectory, of course, makes it hard to capture the value of research, or to have anything like a reliable, predictable path by which scientific research gets turned into marketable products. Exxon’s efforts to develop a practical rechargeable battery ultimately failed, though its research would spawn a successful battery in the fullness of time."
Source: https://www.construction-physics.com/p/how-we-got-the-lithiu...