Technical entertainer; process inventor; living in the Valley of Death; making zillions of tiny switches; #ChBE professor @GeorgiaTech; @Spec__Tech; Ops my own.

Joined July 2012
235 Photos and videos
Pinned Tweet
Thinking we need a Moore's Law for fabs, where every 18 months the number of fabs in the world doubles.
1
8
846
Michael Filler retweeted
And how shall we measure energy? Barrels of oil equivalent, of course! Unless it’s for a car, in which case we shall use the gallon of gasoline equivalent. Or food, where we will use the kilocalorie! But we will call them - Calories. And heating, where we shall use the British thermal unit! And electricity which shall be measured by the kilowatt hour. And nuclear fuel, for which we will use the gigawatt day. And bullets! Whose energy will be measured with the foot-pound. And…
27 Oct 2024
And we pilots shall measure distances in miles! Will that be statute miles or Nautical miles? Both! Visibility will be measured in Statute miles and distance on navigation charts shall use Nautical miles.
9
7
119
10,413
Michael Filler retweeted
13 Oct 2024
Self driving car? > PID loop Rocket that lands itself? > PID loop Hypersonic ICBM? > PID loop Your thermostat? > PID loop
134
286
4,107
294,864
Michael Filler retweeted
Jensen articulates very well a key difference between CPUs and GPUs. Note that NVDA has historically not been the lead adopter of the most advanced process technology. “Parallel processing doesn’t require every transistor to be excellent. Serial processing requires every transistor to excellent. Parallel processing requires lots and lots of transistors to be more cost-effective. I’d rather have 10X more transistors that are 20% slower than 10X less transistors that are 20% faster. They (Intel) were the opposite. Single threaded processing and parallel processing are very different.”
13 Oct 2024
BG2. Ep 17. Double $NVDA! System Level Comp Moat, “Insane Demand”, Inference Explosion 1 B x, Memphis Supercluster, OpenAI, X.ai & more. @altcap @_clarktang @bgurley (00:00) Intro (1:50) The Evolution of AGI and Personal Assistants (06:03) NVIDIA's Competitive Moat (15:51 ) The Future of Inference and Training in AI (19:01) Building the AI Infrastructure (31:35) Inventing a New Market in an AI Future (38:40) The Impact of OpenAI (43:25) The Future of AI Models (46.44) X.ai and Memphis Supercluster (51:21) Distributed Computing and Inference Scaling (55:54) Inference Time Reasoning and Its Importance (01:00:46) AI's Role in Growing Business and Improving Productivity (01:08:00) Ensuring Safe AI Development (01:12:31) The Balance of Open Source and Closed Source AI
1
1
11
979
“[Vertical integrators] don’t face existential risk from a single new technology. Their value is in the system, not in any one of the pieces. A competitor with a better technology would have to build a vertically integrated system around that technology” notboring.co/p/34fec310-e114…
1
129
“[Vertical integrators] take on engineering and system risk, but not scientific or technical risk.” True. But many unforeseen roadblocks on the path to scale likely have a scientific or technical risk flavor.
1
91
“That’s one of the most tantalizing things about Vertical Integrators: there is a clear path to tremendously large outcomes. Not easy, but clear.”
84
Michael Filler retweeted
updated the computing-infused fortune cookies to give the fortune some more personality (also experimenting with a way for you to leave a custom note on the website if you want to gift it)
more computing-infused ceramic explorations.. fortune cookies that give you a new fortune each day
8
13
190
13,903
It’s nutty that late model transistors each cost ~$0.000000001 but a single custom circuit with ~10-100k transistors costs at least ~$3500 and you have to wait 6-9 months.
1
143
Lite Brite. A display that anyone can "code."
1
4
199
We can fabricate this circuit in a single desktop-sized tool in less than a minute 👀
oomf shared this to me and it's pretty wild. Let me introduce you Yokogawa minimal Fab, when everyone is going larger wafers & higher throughput, they make 0.5 in wafer tools all in one form factor, with almost all front-end capabilities, that does NOT need to be in a clean room.
1
1
2
401
“I see a much larger problem at work: Journalists and news executives — and I include myself in this critique — were never equipped with the tools of theory and history to inform self-reflection on our field and to imagine alternative means and models of serving our public.”  Great piece. Also easy to substitute “journalists and news executives” for most anyone and most any discipline. buzzmachine.com/2024/09/11/w…

1
200
Can we discuss the extreme energy intensity of the process itself?
130
Hell’s front porch is PG-13. The pollening is rated R.
It may be false fall, but that doesn’t mean you can’t enjoy it 💛
1
171
Michael Filler retweeted
27 Aug 2024
Vertical Integrators are companies that: 1. Integrate multiple cutting-edge-but-proven technologies. 2. Develop significant in-house capabilities across their stack. 3. Modularize commoditized components while controlling overall system integration. 4. Compete directly with incumbents. 5. Offer products that are better, faster, or cheaper (often all three).
2
2
47
6,275
Mathematicians might want to lawyer up
29 Aug 2024
You may not know that the derivatives of the position vector (x) with respect to time have interesting names: Velocity (v) = dx/dt Acceleration (a) = d²x/dt² Jerk (j) = d³x/dt³ Snap (s) = d⁴x/dt⁴ Crackle (c) = d⁵x/dt⁵ Pop (p) = d⁶x/dt⁶
1
137
“a network connection requires radio communication, through Bluetooth, WiFi, or cellular data. All of these technologies need at least 100 milliwatts of power to run continuously. This isn’t much when you are connected to a mains power supply, but you’ll quickly run down anything battery-powered.” petewarden.com/2024/08/23/wh…

1
854
“Canva for microsystems” 🤔
She was rejected by 100 investors. An outcast among founders. But today, her business is worth $26 billion and has 200 Million monthly users. Here’s her story:
1
175
AI people are the nth people.
26 Aug 2024
The OG Scaling Laws from 1965 Gordon Moore AI people are not the first people to think in terms of log, log plots, they are the second. chiphistory.org/20-moore-s-l…
1
211
We separate almost everything, from fruit to microchips. ~15% of worldwide energy is spent on separation processes. 15%!!
1
1
8
675