I've got enough material I couldn't use for this article to write a sequel. Let me know if you want to hear more about these chemists with their mad schemes.
Love your work, as always. Thank you for clearing the pipes with this fascinating story about how machine learning might accelerate scientific discovery.
I'm interested in reading about delivery on big promises. Such a large grant. Your apt comparisons to this government's technology incubation announcements. Too often the story is "government/grant/whatever announcement...[silence]" as you frequently (& usefully) report. Glad to read about what is getting done.
These "self-driving labs" remind me of the (obvious) comparison with Edison's search for lightbulb filaments. Some 130 years ago, materials development was accomplished by hiring a large team of people to test materials day in and day out. The contrast in terms of acceleration, throughput, and efficiency is striking.
An article on how "brute force" techniques in materials science (or any research domain) have evolved over time would be fascinating. For a recent article contrasting giant mathematical model computation to machine/learning pattern recognition see "The high-tech race to improve weather forecasting" (The Economist, messy hypertext link pasted below).
So, yes, mark me down in favor of a Part II.
My request would be for (continued) emphasis on relatable explanations (I'll make up an example here, "machine learning algorithms are like self-ratcheting straps...under xyz predetermined conditions the software adjusts its pattern recognition criteria in these ways") rather than descriptions that contribute to "AI" hype. It's a tool, and I want to know its limits, what's needed to make/ensure its usefulness, the conditions under which downsides are outweighed by upsides, etc.
For a recent example of the social issues raised by the possibility that humans will over-trust machine output (something that clearly concerns me) see "What can you do when A.I. lies about you?" (NY Times, again, link below).
Thank you very much Paul. This is a great and fascinating portrait of what we are doing with our colleagues, from high-school students to faculty at UofT and our consortium member universities. It is humbling. Thank you for bringing this to your readers! I am going to take advantage of this forum to do an “Ask me anything”. If any of the readers has a question, let me know and I will try to answer it in this thread.
Reporters always regret the questions they forgot to ask. Alán, can you tell us more about the AI that powers these beasts? Is it an off the rack app, or does it require extensive coding? If the latter, what does the coding entail? I got a bit of the answer from your colleague Marty Burke, but I'm still a bit in the dark.
Cheers Paul! The AI that we use is in the fields of "AI for materials and molecules" and also "Bayesian Optimization". We customize the algorithms from these fields to work in our particular applications. We are also developers of many of the first algorithms in this category and keep developing them actively. Therefore, a newcomer could download existing work by us and others in the community to get started, but customization is necessary. I would not call the coding as extensive but as specialized. I could give some further concrete examples.
Hi Alan...thank you for taking the time to be here (and thank you for coming to Canada to build this project!). My question may seem absurd after having received $200 million, however, I'll ask it nonetheless. What else do you need to realize the full potential of the work AC is doing - are there gaps that you can see (whether in the private sector or in the public sector)?
Ben, so many things! First of all, the 200 million dollars comes with many strings attached. For example, we don't have enough money for equipment, so we are all writing extra grants and external projects to fund the equipment. Second, there are always applications that were not envisioned in the original project, and that is why we are working to attract industry to build self-driving laboratories for specific problems for their interest. The building that we invested on needs to be named, so hopefully philanthropy can help UofT recover the investment in the construction of the building. We would love to have also philantrophic donations for new chairs in the area so that we can bring new professors from Canada and abroad to Toronto to cover new exciting areas. In a sense, we need to become self-sufficient after the seven years of federal investment and that needs attracting all sorts of people to care, participate, and fund, aspects of the project.
The optimist in me - having some early exposure to high performance computing - has believed that the combination of smart people and greater computational heft would start making big problems look quite small. Back then, Queen’s researchers saw weeks of waiting for results crunched into hours. Twenty years later, the promise is exponentially greater. People are smart. And our tools are getting almost as smart.
And don’t overlook the significant implication buried in the details. This person would have been welcomed with open arms anywhere, and chose to relocate here because of greater social peace, which is also deserving of investment.
Wow. Like, really WOW!! A guy in a univeristy that is not obsessed or oppressed by the latest sociological fantasies. And who would have heard of him, were it not for reporting like yours.
Great reporting. Kudos to U of T, CIFAR and the federal government for this joint effort. Public funds well spent. And here’s to Alan Aspuru- Guzik. He deserves our support and our respect.
Thanks for this Paul. Your time and our subscription fees are also well spent!
This on the ground report reminds us of the potential of human discovery and how research is a badly needed counterpoint to the constant doomsday narrative that emanates out of the climate alarm industry.
I’m pleased that the Canadian Government is providing grant money to develop the AI Lab potential. It would be quite an achievement if the U of T produced an organic battery that didn’t require heavy metals (and made from bitumen)! Or found an additive to introduce to tailing ponds that turned the pollutants into carbon sinks. This is just the tip of the iceberg and Canada needs to nurture the concept and build a AI Lab supercluster that can rival counterparts anywhere in the world.
This is an excellent example of Canada's ability to play at a top-tier level in knowledge diplomacy. And who knew that U of T is ranked 2nd in North America among public universities. That's huge.
Wow now that is something. I need to re read it a couple or more times to get it into my head. Full disclosure I found Chemistry and Organic Chemistry really hard. Need the sequel for sure.
I've got enough material I couldn't use for this article to write a sequel. Let me know if you want to hear more about these chemists with their mad schemes.
Hell, yeah. The "Ages" are defined by their materials science! (Stone, Bronze, Iron)
This is the real news.
Yes, please. More. Optimistic and Inspirational!
Yup, yup, yup and yup.
Yes yes yes
Yes please.
Paul,
Love your work, as always. Thank you for clearing the pipes with this fascinating story about how machine learning might accelerate scientific discovery.
I'm interested in reading about delivery on big promises. Such a large grant. Your apt comparisons to this government's technology incubation announcements. Too often the story is "government/grant/whatever announcement...[silence]" as you frequently (& usefully) report. Glad to read about what is getting done.
These "self-driving labs" remind me of the (obvious) comparison with Edison's search for lightbulb filaments. Some 130 years ago, materials development was accomplished by hiring a large team of people to test materials day in and day out. The contrast in terms of acceleration, throughput, and efficiency is striking.
An article on how "brute force" techniques in materials science (or any research domain) have evolved over time would be fascinating. For a recent article contrasting giant mathematical model computation to machine/learning pattern recognition see "The high-tech race to improve weather forecasting" (The Economist, messy hypertext link pasted below).
So, yes, mark me down in favor of a Part II.
My request would be for (continued) emphasis on relatable explanations (I'll make up an example here, "machine learning algorithms are like self-ratcheting straps...under xyz predetermined conditions the software adjusts its pattern recognition criteria in these ways") rather than descriptions that contribute to "AI" hype. It's a tool, and I want to know its limits, what's needed to make/ensure its usefulness, the conditions under which downsides are outweighed by upsides, etc.
For a recent example of the social issues raised by the possibility that humans will over-trust machine output (something that clearly concerns me) see "What can you do when A.I. lies about you?" (NY Times, again, link below).
Yours,
Vince
https://www.economist.com/science-and-technology/2023/07/26/the-high-tech-race-to-improve-weather-forecasting
https://www.nytimes.com/2023/08/03/business/media/ai-defamation-lies-accuracy.html?unlocked_article_code=Au_RuE4diGTaPPm653vQDSIEAZcwTPTgw2NxwBUwD820SSPeSd9rmc6AD8PXC2qbb6ix1SLpH8bZJcEA9rG6MyVYJmOCwfvmXLYlZwNjcSqmMlxJ77h9v-u4x41SWf4H_ccwrMyM_A5oFIFMyU0v99lgtZgIB3NTgobPqgUrt4ldqAHvHIhG8fxcattE9jXXgLED76OTahj4cmNWoS_32uhXLSCSZCrV-HMmW8Z578XuBKT2zh8ur2kmoxuDTLZZFO4uTf3Sl08wbUjxNzxB4_nNWc0EFNBSpivHKe9NClQItMHjNyOpSB8oaWnt4FrmkNjgQ2oM-Dez8exNZWoIaPrIP-fAAuEmjCd6ig&smid=url-share
Well thanks for really and truly over delivering on your promise. Great piece. Sequel please.
yes, more please.
Encore!
Yes please!
Definitely yes!
yes please!
Please! Great read.
Thank you very much Paul. This is a great and fascinating portrait of what we are doing with our colleagues, from high-school students to faculty at UofT and our consortium member universities. It is humbling. Thank you for bringing this to your readers! I am going to take advantage of this forum to do an “Ask me anything”. If any of the readers has a question, let me know and I will try to answer it in this thread.
Reporters always regret the questions they forgot to ask. Alán, can you tell us more about the AI that powers these beasts? Is it an off the rack app, or does it require extensive coding? If the latter, what does the coding entail? I got a bit of the answer from your colleague Marty Burke, but I'm still a bit in the dark.
Cheers Paul! The AI that we use is in the fields of "AI for materials and molecules" and also "Bayesian Optimization". We customize the algorithms from these fields to work in our particular applications. We are also developers of many of the first algorithms in this category and keep developing them actively. Therefore, a newcomer could download existing work by us and others in the community to get started, but customization is necessary. I would not call the coding as extensive but as specialized. I could give some further concrete examples.
Hi Alan...thank you for taking the time to be here (and thank you for coming to Canada to build this project!). My question may seem absurd after having received $200 million, however, I'll ask it nonetheless. What else do you need to realize the full potential of the work AC is doing - are there gaps that you can see (whether in the private sector or in the public sector)?
Ben, so many things! First of all, the 200 million dollars comes with many strings attached. For example, we don't have enough money for equipment, so we are all writing extra grants and external projects to fund the equipment. Second, there are always applications that were not envisioned in the original project, and that is why we are working to attract industry to build self-driving laboratories for specific problems for their interest. The building that we invested on needs to be named, so hopefully philanthropy can help UofT recover the investment in the construction of the building. We would love to have also philantrophic donations for new chairs in the area so that we can bring new professors from Canada and abroad to Toronto to cover new exciting areas. In a sense, we need to become self-sufficient after the seven years of federal investment and that needs attracting all sorts of people to care, participate, and fund, aspects of the project.
I feared as much. Perhaps some folks reading this will be encouraged to get involved...
Thanks again!
Of course, happy to talk to people that want to get involved in supporting the project. Thanks for your question.
Okay. you made my day with this most excellent piece of work. Thank you thank you thank you!
G*d damn. This might be the single most inspiring thing I've read all year. Thank you Paul and thank you Alan.
Paul this wonderful story is a good example of why you are special.
By all means.
The optimist in me - having some early exposure to high performance computing - has believed that the combination of smart people and greater computational heft would start making big problems look quite small. Back then, Queen’s researchers saw weeks of waiting for results crunched into hours. Twenty years later, the promise is exponentially greater. People are smart. And our tools are getting almost as smart.
And don’t overlook the significant implication buried in the details. This person would have been welcomed with open arms anywhere, and chose to relocate here because of greater social peace, which is also deserving of investment.
I very much want to read more. Stories about motivation for change linked to ambition are sorely needed nowadays.
Wow. Like, really WOW!! A guy in a univeristy that is not obsessed or oppressed by the latest sociological fantasies. And who would have heard of him, were it not for reporting like yours.
Great reporting. Kudos to U of T, CIFAR and the federal government for this joint effort. Public funds well spent. And here’s to Alan Aspuru- Guzik. He deserves our support and our respect.
Thanks for this Paul. Your time and our subscription fees are also well spent!
This on the ground report reminds us of the potential of human discovery and how research is a badly needed counterpoint to the constant doomsday narrative that emanates out of the climate alarm industry.
I’m pleased that the Canadian Government is providing grant money to develop the AI Lab potential. It would be quite an achievement if the U of T produced an organic battery that didn’t require heavy metals (and made from bitumen)! Or found an additive to introduce to tailing ponds that turned the pollutants into carbon sinks. This is just the tip of the iceberg and Canada needs to nurture the concept and build a AI Lab supercluster that can rival counterparts anywhere in the world.
An awesome look at a difficult to grasp subject, well done!
This is an excellent example of Canada's ability to play at a top-tier level in knowledge diplomacy. And who knew that U of T is ranked 2nd in North America among public universities. That's huge.
This is good news, public funds being well invested in this inspiring adventure in chemistry.
We need more of this story. Thank you.
Superb piece!
Wow now that is something. I need to re read it a couple or more times to get it into my head. Full disclosure I found Chemistry and Organic Chemistry really hard. Need the sequel for sure.