The Ledger
Every statement that passed quotation and attribution checks. Mix any filter with any other: certainty 1/5, debate potential 5/5, or both at once.
why aren't all 2,374 resolved? a statement only gets an assessment when the public record can support or contradict it. opinions and what-ifs never can, and 6 checkable ones are still open, waiting for their date. predictions held up or didn't; assertions are supported or contradicted. on every card: ▮▮▮▮▮ certainty · ▮▮▮▮▮ debate potential. speakers are clickable
Zhou: Lamini cuts LLM fine-tuning time from months to milliseconds
“And by efficiency, I mean, you know, it's instead of something that might take weeks or even months that's bringing it down to even like the millisecond level.”
Zhou: Lamini is the only platform running LLMs on AMD GPUs
“We are the only folks who can actually run your language models on top of AMD AMD GPUs.”
Lamini's AMD support unlocks 20,000 GPUs, enough to train GPT-4
“So that unlocks, what that means is that unlocks about 20,000 GPUs readily available today for enterprises to be able to use, and to get a sense of what that means you can train GBD-IV.”
Zhou: Outsourcing specialized medical AI data labeling to Scale AI failed
“We tried outsourcing actually to like scale AI, et cetera. None of that worked. It had to basically be me.”
Zhou: LLMs can reach 99% accuracy today with narrow scoping
“I think we can get to that performance today, but it's based on how you scope out the problem. So if it's a very narrow scope, of course you can get that.”
Lamini has achieved software parity on AMD GPUs with CUDA
“We have reached software parity with essentially CUDA.”
Srinivas: Open-source AI models do not match closed-source model capabilities
“That needs the open source models or fine tuned versions of them to match the closed source, the closed ones. And That's not the case today.”
Srinivas: Self-hosting open-source LLMs is more expensive than closed-source APIs
“Even if there exists a checkpoint today, that's as good as the close models. It's still not an incentive to use them and serve them yourself because you're going to burn more money than paying for the close ones because the close ones are being served in such …”
Shah: Patient non-adherence is healthcare's biggest problem, not diagnosis
“Our biggest issue in the U.S. And most of the developed world is the people don't follow the directions after they're diagnosed. Like, it's not an issue of diagnoses. It's an issue of adherence.”
Shah: Foundational LLMs are instruction-tuned for queries, not multi-turn conversations
“They've all instruction tuned for what I call a query, not a conversation.”
Socher: Salesforce research team invented prompt engineering in 2018
“As part of the research team, we actually invented, ah, prompt engineering in 2018”
Biewald: Most enterprises have not deployed LLMs into production yet
“I think that LLMs in particular, we talk to a lot of the people and we don't see a ton of people getting them into production yet. And I think it's funny, like VCs are always surprised, like when we tell them that I think that I don't know. I'm bullish on LMS,…”
Biewald estimates 99% of Global 2000 use ML for core operations
“I bet 99% of the global 2000 is using machine learning for something that they actually really care about.”
Ratner: $200 of ChatGPT calls can clone closed models into open ones
“If you take a couple hundred bucks of API calls to, say, ChatGPT, and you graph that onto a model that is substantially smaller, say a seven billion parameter model like LLAMA, or now increasingly fully open for commercial use ones, like Red Pajama is one that…”
Ratner: Distillation creates specialized models 100,000x smaller with higher accuracy
“It shows that you can take these massive kind of generalist foundation models, And not only can you tune them using that kind of programmatic labeling to be more accurate on a given task that you need to be really accurate on, but you can then distill down ver…”
Retrieval-augmented generation is 100 times cheaper than fine-tuning, says Amr Awadallah
“The grounded generation approach is real time. A new fact comes in is showing up in the answers right away. There is no hallucination. I mean, you minimize it significantly and the cost is a hundred times cheaper.”
Turck: Cost pressures accelerate architectures using S3, Iceberg, and DuckDB
“And then there's the acceleration of like alternative architectures, you know, storage. Why don't you just dump everything in S three, which is very cheap. And then, you know, you could use iceberg on top to add some structure to it. You can, we talked about D…”
Turck: Google invented the Transformer but is playing catch-up
“So in GPT, which stands for generative pre-trained transformers, the T is a transformer architecture that was actually Developed at Google, and the irony is that Google is playing catch-up to Microsoft and OpenAI right now because OpenAI accelerated the develo…”
Turck: Academia has completely lost the AI research battle to industry
“Academia as of now has just completely lost the battle. In AI research and development, the darker dots, and this is a long time scale at the bottom, show that in the early years of the current wave, a lot of the research came out of academia, but today it's p…”
Reis: Data engineering paradigm is swinging back from ELT to ETL
“But what's interesting right now is the discussion is actually moving back to ETL. I, I'm starting to see more and more discussions about how ELT is a bad paradigm, and again, remember when I said talk about pendulums? The pendulum again is swinging back to ET…”
Handy: Early-stage VCs doubted data practitioners would learn Git in 2016
“Early stage VCs that I spoke to back in 2016 told me that, like, it wasn't at all clear that data practitioners actually wanted to learn Git.”
Azhar: Companies ignoring AI and automation shed jobs, while adopters grow
“That the data seems to show is that companies who invest in robotics, automation, AI grow. But it's companies that don't that shed the job.”
Eifrem: Only 5 modern NoSQL database companies have crossed $100M ARR
“There's five database companies that have crossed a hundred million, right, of the, kind of the, let's call it the NoSQL crowd, or like modern operational database companies. It's, You know, Mongo, and then it's us and Redis. We're on that kind of Mongo path. …”
Eifrem: Neo4j is frequently 1,000 times faster for connected data queries
“We've optimized every layer in the stack of the database architecture completely around connected data. We're not built on top of a different database or anything. It's a native architecture. And that means that if you want to query along how things are connec…”
Craib: Numerai has more data scientists than any other hedge fund
“And so we have the best modeling talent because we have much more More, many more data scientists working on Numeri than any other hedge fund in the world.”
Dehghani: Current Data Mesh setups are Frankenstein creations stitching existing tech
“At this point in time, it looks like a Frankenstein creation because we have to stitch together a lot of technologies that exist and they weren't designed for this model of reconfiguring, you know, being reconfigured in this way.”
Ghodsi: Early Big Tech achieved AI breakthroughs using 1970s algorithms with massive data
“What they were doing is they were taking those algorithms from the seventies that do not work, but they were applying orders of magnitude, more data to it. So a lot of data on modern hardware, and they were getting superhuman results.”
Ghodsi: Snowflake coexists with Databricks in about 70% of accounts
“It's certainly going to coexist, and it already coexists with Databricks in probably 70% of the accounts we're in.”
MongoDB replaced six C-level executives in the three years post-IPO
“It's a, it's not a secret, but since we've gone public three and a half years ago, we've had six C-level changes, right?”
95% of machine learning models at Google and Facebook are neural nets
“So if you look at a company like Google or Facebook, almost all of their models, like, 95% of their models are neural net based.”
Narayan: Materialize provides 99.9% of batch database functionality
“And we really think materialize the product is the first database that gives you know, I don't want to say literally all the functionality, but to the morally speaking, you know, 99.9% of the functionality that you can write in a batch database.”
Stewart: Cockroach Labs adopted BSL, open-sourcing code after three years
“One of the big things we did is we did change our open source license. So it's still source available. It's still free, but we moved from Apache to a license called the business source license. Which is just like Apache, you can use the software, you can redis…”
Stewart: CockroachDB beats Amazon Aurora heads-up in single data centers
“We're still winning heads up versus Amazon Aurora in single data centers.”
Pesenti in 2020: Multi-million-dollar AI training runs are unsustainable for Facebook
“Where it becomes millions is when you do training runs. So some of the training runs in the most advanced system that it comes from our company or other companies out there are starting to be extremely expensive. Yeah. Like you can look at one run in the scale…”
Elsamadisi: Narrator reduced 40-hour Looker queries to just 10 minutes
“We've taken queries that were taking, like, literally 40 hours in Looker, and we've dropped them down to, like, 10 minutes. Queries that cross billions of rows end up happening in, like, seven to eight seconds on a warehouse.”
Elsamadisi: Redshift, Snowflake, and BigQuery perform identically on long tables
“The speed improvement on Redshift and versus Snowflake versus BigQuery becomes interchangeable. Because what they handle is for those edge cases, we have like a bunch of, like Snowflake is really good at a bunch of small tables everywhere trying to load a memo…”
Volpi: AWS launched DocumentDB to bypass MongoDB's SSPL license
“Mongo took an interesting similar approach. They have something called SSPL. In that case, it says you can take it, but any software that you build around it to run it, you have to open source it. So Amazon will have to open source its entire software base use…”
Volpi: Google and Azure partner with OSS creators to counter AWS
“And then the other thing that you do As a business is you partner with Google and Azure who, because they're pretty far behind Amazon, they're more than happy to have the original item and to pay you something for the privilege of running your version.”
Volpi: Employees write 95% of code in open source startups
“Nowadays, most of these open source companies, you know, 95% of the code is written by employees, and the five percent are like patches or connectors or something like that that the community is building.”
No living neuroscientist can explain what a thought is, says Thomas Reardon
“I'll tell you there's not a neuroscientist alive who can tell you what a thought is.”
CTRL-Labs reconstructs motor neuron activity before physical muscle movement occurs
“We figured out how to actually reconstruct the activity of the motor neurons that drove that electrical activity in the muscles themselves. This is activity that actually happens long before The muscles actually move, you know, long in the computational sense …”
Reardon: CTRL-Labs captures single neuron signals without penetrating the skull
“This is a, Pretty significant breakthrough that we're able to get to single neurons without basically penetrating into your skull.”
A trained CTRL-Labs band will not work at all on another person
“This is a very personal model that is custom to just you. When you put on the band after you've trained it, if you put it on somebody else, it doesn't work at all. It is, and it's not like design. That's just the way neural interfaces are required to work. The…”
CTRL-Labs uniquely identifies users among 7.4 billion in 800 milliseconds
“I can put a band on you and after 10 minutes I will know enough from those signals that I can pull you out of a population of 7.4 billion people after 10 minutes of training data and do that in about 800 milliseconds.”
Marcus: Teslas crashed into multiple emergency vehicles due to vision flaws
“It turns out that Teslas have a problem with stopped vehicles, mostly emergency vehicles, on the side of the road. So they have run into, in the last year, three fire trucks, a tow truck a police car, and so forth.”
Uretsky: Kelsey Hightower rated DigitalOcean's Kubernetes UX superior to Google's
“Kelsey Hightower, one of the core contributors to Kubernetes and over at Google, our friendly company, actually surveyed a bunch of different providers delivering Kubernetes services and said that DigitalOcean provides the best developer experience And is the …”
Gourley: Information propagation in networks is central to Russian military strategy
“And it's important to know the context of where information came from, because as information is seeded into the networks, it starts to propagate, and this becomes key attack vectors that are actually unfolding, and if you listen to some of the major Russian g…”