Current LLM architecture is so inefficient that most people don’t realize how fast GPUs are.
A *single* token generation for 1T model spends same number of flops as
- simulating 2hr weather for small city
- chess engine playing half game
- multiplying two 255000 digit numbers
We are basically brute forcing our way to AGI.
Many people think any given ML project is 99% training.
In reality, it’s 50% evaluation, 40% data cleaning, 8% integration, and 2% training.
The first two set the noise floor for learning. No ML magic matters; the model cannot lower the noise floor, as that’s the optimal bound of Shannon encoding of your data.
Thus, not a single day goes by without me thinking about ontology. Even the old labels have to be constantly reviewed.
🇨🇳 One of the world’s largest container ships began its first voyage from Shanghai, and the scale is insane.
The CMA CGM Notre Dame is 400 meters long, can carry more than 24,000 containers in one trip, and was built in China, fueled in China, and launched from China.
It runs on cleaner LNG fuel and uses AI to read weather and ocean currents, choosing safer and faster routes across the Asia-Europe trade corridor.
Source: ShanghaiEyeMagic on YT / Writer: Sol
Following a very difficult meeting with my accountant, I just found out how much it is going to cost me in terms of an "exit tax" to leave Quebec and Canada. No human being in a free society should have their hard-earned money stolen in this manner. I'm genuinely numb. I'm speechless.
@Zai_org this trained on Huawei. We may see western software companies eventually training on Chinese hardware. Very similar to how it is now, west brands built on Chinese manufacturing.
B2B cold outbound is dying and I don't think people realize how fast.
I get 30 spam calls a day.
My inbox is full of newsletters I never signed up for.
Gmail is getting smarter. Carriers are getting smarter. AI is going to compress your entire inbox into a summary you skim in 10 seconds.
Cold email and cold calling will be irrelevant within 2 years.
B2B companies are going to have to run consumer playbooks. UGC creators, Meta ads, YouTube, paid social.
The companies still building pipeline on cold outbound are building on a foundation that's already cracking under them.
The blue ocean strategy is risky.
It is expensive.
You are often trying to force a round peg into a square hole.
Blue ocean folks will spend months or years trying to invent something new based on a theory when there's no proven demand.
I’d rather swim in a red ocean where I can pick an existing product/service with a proven business model in a proven market that's making money today.
It's much easier and becomes profitable faster with way less risk.