Completely a coincidence, but this morning (9-10 am), I gave a Tepper Faculty Coffee Seminar on Unconventional Computing. This was not recorded.
Sometime in January, having just returned from a wonderful trip to UCL and Cambridge (see Lean Monarchy) where I spoke about Quantum Integer Programming, I received this email:
Dear Professor Tayur,
I was wondering if you would be interested in giving a talk at TEDxCMU this year in our annual conference on March 28th, 2020 at McConomy Auditorium. We were very impressed with your work in the quantum computing group and think our audience would be very excited to hear what you have been working on.
The TEDxCMU team works year round for this conference and speakers at our event include people like Freddie Anzures, a designer who worked with Steve Jobs on the first iPhone, and Jane Metcalfe, the co-founder of Wired Magazine. The talks given at our event are then published by the official TEDx YouTube channel.
We hold one or two seats for the faculty of Carnegie Mellon and I think you would be a great fit. Let me know what you think!
At that time, the plan was for the event to be “in-person, on-stage” (yes, an over-rehearsed monologue) with some eye-popping background slides, the standard TED format.
I was hesitant to accept.
Why? Because I find TED talks, in many cases, to be inauthentic in addition to being intellectually suspicious (if not outright rubbish, especially in the social sciences, in light of the replication crisis). More a staged performance of memorization with practiced spontaneity.
This was not first time I have been invited, having twice previously declined to participate.
Not to mention that suffocating sameness that is imposed on the speakers (through an approval process that is in itself an unsavory experience, and then expected to deliver a rehearsal of that approved script, on stage, the day before), a cookie-cutter production, the mass-manufacturing of Stepford Wives, creating a synthetic homogeneity (cloning Agent Smith?), stamping out all individuality and snuffing out the joys of authentic spontaneity.
TED talks are maximally inverse to what I would want to do!
For instance, I thoroughly enjoyed presenting an unscripted, unrehearsed monologue on quantum computing (in 2018), which has been received very well, a recent email being:
I am the former CTO of the U.S. Department of Energy, am involved with the National Quantum Initiative via the Office of Science (inside the DoE), and just finished watching your “Moonshot: Quantum Computing” video on YouTube. Simply put….awesome, entertaining, enlightening.
I would welcome an opportunity to chat with you.
And I have previously mentioned a most startling email in Now to Wow.
There is a “but” coming about TED, you can feel it!
On the other hand, having said all of the above, I do recognize that:
TED is a global brand, with a well-oiled machinery, a platform with immense reach, a curated channel for video content distribution like no other.
Furthermore, viewing the talk as a teaser, a preview, an invitation to the audience to learn more, especially about something whose very existence they might have been entirely unaware of until then, and not as an end-product, I found myself warming up to the idea.
Channeling Renee Zellweger from Jerry Maguire, I justified:
You had me at official TEDx Youtube channel. 😏
The one-time prep and delivery, however unpleasant to me to conform (to such superficiality and be forced to follow an “approved” script), I felt, was a worthwhile “cost” for what I really thought was the “benefit”:
A “forever” video on a global platform that can serve as an invitation for folks to learn more.
I was feeling meh, resigned to this self-inflicted fate, the rational calculus of “benefit being over cost” notwithstanding.
I could hear the smug Agent Smith from The Matrix:
Do you hear that, Mr. Anderson. That is the sound of inevitability. That is the sound of your death.
Then came COVID19.😳
Event cancelled entirely? No. Do it via zoom? Yes. Live telecast? No. Record it on zoom, but not the over-rehearsed inauthentic uninterrupted speech? Yes. Home-production with TED post-production? Yes. Q&A format? Yes.
There is another “but” coming.
However, I did not want to do a garden-variety quantum computing talk, as there are so many of them out there, narrowly fixated on superconducting qubits and cryptography.
If the format is going to be so radically changed, then, I wondered, why not also change the content? Yes.
Now, finally, I was actually excited about TEDxCMU 2020! In my maximally inverse framing:
COVID19 breathed life into my TEDx prep!
What topic then?
Something I felt was not as widely publicized as it should be, an idea worth spreading:
More magical than quantum computing.
Back to The Matrix:
Goodbye, Mr. Anderson.
Here it is, with inauthenticity stripped, yet articulating the fancifulness (and miraculousness) of the given object that is the chosen subject of my TEDXCMU talk:
To keep within 8-minutes, and to not get too detailed on a particular question, some of the content was not filmed, and a few of my answers were edited. The “full transcript” is reproduced here.
Q: What do you mean by unconventional?
A: I mean anything that is not what we use in our laptops, iphones, ipads, or in the servers in the data centers. Something that is not classical computing, anything that does not rely on the manipulation of bits.
Q: Is this different from quantum computing?
A: Quantum computing is just one example of unconventional computing.
There are many different types of quantum computing. Regardless of which type, the core commonality is that we are manipulating superpositions of zeroes and ones, and not “already formed” zeroes and ones. These are called quantum bits, or qubits. They are very different from bits. This is the quintessential feature of quantum mechanics that has no classical counterpart. This makes quantum computing unconventional.
Q: What are the different types of quantum computing?
A: Let me walk through four of them.
First: Circuit (or Gate) based, the ones you hear most about in the press because Google and IBM are investing heavily in it. This is “conventional or mainstream quantum computing”. Here too there are different types: (a) superconducting (Google, IBM) and (b) ion traps (by a company called IonQ).
Second type: Measurement based, also called one-way quantum computing.
Third type: Topological quantum computing, that Microsoft is attempting.
Fourth type is Ising based, sometimes referred to as Adiabatic Quantum Computing, or quantum annealing. This is being developed by D-Wave (in Canada).
Q: What are other non-quantum examples of unconventional computing?
A: We have been so accustomed to classical computing with our laptops and iphones, that is digital, that is based on transistors and integrated circuits that manipulate zeroes and ones using Boolean logic gates, that we have largely forgotten, actually overlooked, an entirely different way of calculating: analog computing.
An example of non-quantum analog computing that is unconventional is something that uses lasers is called Coherent Ising Machine.
All of today’s mainstream computing is digital-classical. What is exciting is to pair it with something that is analog-unconventional.
It is time to rethink, reimagine and reinvent computing, and look beyond Turing machines and von Neumann architectures.
Q: What attracted you specifically to Ising based computing?
A: Ising computing is unconventional even within the unconventional!
Pragmatically speaking, we already have lot of ways of doing Ising computing. These use optical components and digital technologies available today and can work at room temperature. They can be miniaturized and likely will be far less expensive.
I suppose you can call me a pragmatic futurist!
Q: How do Coherent Ising Machines (CIM) work?
A: Very differently from conventional computing, and also very differently from quantum computing.
We are manipulating lasers with electronic signals, and, after a few “photon-electron” interactions that we can program, the light wave bifurcates, which can be interpreted as a zero or a one, and the desired solution emerges!
It is magical!
So CIM is not manipulating zeroes and ones, either in bits (like conventional computing) or qubits (as superpositions of zero and one), like quantum computing. CIM is electronically perturbing light waves.
NSF just supported research at Stanford and NTT (Japanese firm) with a $10 Million grant to develop CIM.
Q: What can we do right now with unconventional computing?
A: Lots of things! Break down unconventional computing into two parts: (1) unconventional algorithms and (2) unconventional hardware (like CIM or IBM and Google’s machines).
There are many Ising approximations — classical algorithms — available today that we can run on classical machines (like our laptops) as well as GPUs. We have developed, based on Ising model framework, super-fast classical algorithms for special structured problems. These can solve hard problems in finance, supply chain and cancer genomics right now. These unconventional algorithms are already impactful even before we actually have unconventional machines, which will bring further speedup.
Q: Can you elaborate on a specific example?
A: Let us look at global supply chain problems from the chemical industry. We have to optimize several decisions, simultaneously, regarding suppliers, manufacturing, storage and transportation with considerable variation in prices and demands over time.
These are very time-consuming to solve not only because the models have many variables, but because these variables are binary (zero or one), and the objective function is non-linear in very unmanageable ways.
Solving these models today using classical algorithms can take hours, and even then, in many cases we do not obtain optimal solutions either because we are running out of space, or it is running way too long. We may be leaving a lot of money on the table, so to speak.
We observed that while the objective function is indeed horribly complicated, the constraint matrix has structure particularly well suited for a solution method that is Ising-based. This is the breakthrough contribution, as until we came along, there was no really new practical quantum-inspired algorithm that tackled real-world sized problems.
We call this Graver Augmented Multiseed Algorithm, GAMA.
This is a completely new invention because we are now thinking unconventionally. We are able to solve many instances 100 times faster, and tackle problems many orders of magnitude larger. Right now.
Q: What do you see as future possibilities with unconventional computing?
A: Classical computing is not going to go away anytime soon. The future will be a hybrid between classical and unconventional computing. We do not have to wait for decades to have fault tolerant quantum bits (qubits), in sufficient numbers and connectivity, as can get a great boost with Ising computing before that.
Many nations (notably Japan, China and many in Europe) have taken notice to these new possibilities. Indeed, not to be left behind, the US Government has already budgeted “workforce development” in this area to prepare the next-generation of knowledge workers.
It is indeed an exciting time to be doing unconventional computing.