|
Dear QLeapers & Friends,
Hope the semester is winding down nicely for you...
Two years ago, a former business partner of mine in the Performing Arts industry asked me, knowing that I had become fully embroiled in the world of Quantum & AI, "How can a 'Next-Word Predictor" ever be creative?"
By "Next-Word Predictor", he was referring to LLMs, the workhorse of today's AI. And that description has apparently been actively debated for some time, with well thought-out arguments ranging from
German philosopher Ludwig Wittgenstein once said: "The limites of my language are the limites of my world." I bet many Comp. Sci. students will get what he meant just as readily as many of their counterparts in Liberal Arts.
Yes, as a mechanism, an LLM's perceivable output is driven by prediction of the next token. (NB: A token is not a word, but rather a chunk of text -- somewhere between a character and a word -- that an LLM uses as its basic unit of processing.)
But it usually uses a vocabulary with up to a hundred some thousand tokens — each embedded in tens of thousands of dimensions, and connected through trillions of parameters — to encode, map and correlate the totality of human knowledge as expressed in written forms. (NB: The post-LLM models seek to cover the same in other forms, such as spatial, visual, olfactory, etc. etc.)
That's when this spooky thing in our universe called "emergence" happens, which has gotten us where we are today, in a nut-case...uh, sorry...nut-shell, so to speak.
A year ago, I heard a senior academician jokingly say that you need 3G's to do research nowadays -- Grants, Graduate students, and... GPUs!
Which did not surprise me much, since two of the 2024 Nobels (physics and chemistry) were already maximally "AI-entangled".
Yes, it seems that today no matter where you roam, AI is the elephant in the room...
But How About Inside The Ivory Tower?
Harvard, UT, Yale, Stanford, and Princeton, the 5 US universities with the largest endowment funds, currently register ~$200Bn in their collective coffer, whereas Nvidia, Alphabet/Google, Apple, Microsoft and Amazon, the 5 largest US corporations in market cap, together clock in at ~$20Tn.
That's ~100x of the academic group. For comparison, in the early 90's, the gap between the first group (the same 5 schools, at ~$20Bn) and the second cohort (a different set, at ~$400Bn) was only ~20x.
Here's a simpler correlation to take home: Bill Gates' net worth in 1995, when he first topped the "World's Richest" list, was $12.9Bn, on par with the 5 schools' total at the time. Today Elon Musk leads the pack at ~$682Bn, almost 3.5x the current total of The Five.
Of course, Harvard can't just spend its $52Bn endowment, and neither can Nvidia spend its market cap of $4.8Tn. But here's the twist:
- Back in the day, any one of those 5 universities would have been able to buy the hardware required for any type of frontier R&D they chose to pursue.
- Today, none of them could afford the enterprise AI-training GPUs to just be on equal footing with the likes of Open AI or Google DeepMind. Each of those leading industry AI labs has amassed (or is amassing) a million or more of those chips under their belt. At $30-40K a pop, those chips alone would cost tens of billions of dollars. And we're not even talking about the additional billions needed to cover wages, electric bills, and RAM (another chip crunch story altogether!)
So Is There Simply No Way Out Then, You Ask?!
Well, believe it or not, quantum may come to the rescue at some point. One of the ongoing developments in the Convergence of Quantum and AI is dubbed "Quantum for AI" (Q4AI), where quantum or quantum-inspired algorithms could significantly reduce AI's computational expenditure, which is usually embodied by those very GPUs and RAM involved. Another more recent development points to more radical paradigm shifts where classical neural network nodes (and GPUs by extension) are replaced with "thinking" quantum particles.
In fact, I had the opportunity and pleasure to help either lead or shape those very discussions at a series of events on or around April 14, aka the World Quantum Day, including...
1. PNW's inaugural Quantum Symposium during its Days of Discovery program, an annual flagship event showcasing student research. Keynoted by CEO of Chicago Quantum Exchange, Dr. Kate Timmerman, the Symposium brought together nearly 20 presenters across the region, from both academia and industry.
Particularly noteworthy: PNW was likely graced for the first time by not one but two hands-on quantum devices which can be used as a toy quantum computer for teaching. One was the Quantenkoffer ("quantum in a suitcase"), a room-temperature photonic system by a German company — the same device MIT reportedly paid $0.5Mn for. The other, built by local startup QuantCAD, required cryogenic cooling. Fortunately, the team brought their own liquid nitrogen.
 Two hands-on quantum devices being demo'd at the QS (Photo album)
2. A Quantum AI Convergence panel @ Union League of Chicago Club, which was part of an event series designed to equip the broader business community with practical quantum literacy across the entire Quantum Information Science and Technology (QIST) spectrum so they can begin envisioning quantum‑driven applications relevant to their sectors.
A key signal from the discourse that evening was nicely summarized in an attendee's (Krystal Rogers) LinkedIn post following the event. She wrote: "The impact will show up inside industries, not as a product you “use”. Healthcare, energy, materials, and finance will be the early innovation recipients, with value showing up in better outcomes, not a new app on your phone." 
What serendipity -- our name is on the title slide... No, it was not staged! (Photo album)
Inspired, or at least intrigued?
Then you may look into two related NEW TECH / GEN ED courses in the fall:
- Our Quantum Future (CS 20010), a quantum literacy course, and
- AI In Practice (CS 22110), an AI competency course
Speaking of which, what better place to use a Case Study here: When news about the rescue of the second F-15 airman downed in Iran first broke out in early April, some early online posts cited secret tech involving quantum magnetometry and AI. And our OQF class happened to be learning about NV-center diamond sensors, the very heart of the subject matter in concern!
Please note there is NO prerequisite to either course, NOR will there be any coding or math involved. Yet both help prepare you for the future, regardless of your career path chosen...
Looking to earn some cash this summer?
We're seeking two enthusiastic student assistants to help bring quantum concepts to life for high school students with hands-on activities, classroom projects, computer labs, and outdoor games. July 20–24, 8:30 AM–3:30 PM. Interested applicants can complete the following form by Friday, May 8th https://forms.gle/G3UmBAXDfNUgrUW96.
Last but not least, best luck on your finals and big congrats to the grads!
Entangledly Yours,
Shawn
|
|
|
|
Thanks for this email!