I spent this week working in Helsinki and I love the city, each time I visit I am surprised by something new, awesome or indeed random. This visit has been no exception discovering the last of the Finnish Air Forces MIG-21 BIS fighter jets capable of Mach 2 and last flown in 1998 parked on top of a building
#BitterLesson scalable learning approaches powered by computation will beat general methods like scaling model size and leveraging vast data sets. Source Jim Fan
https://openai.com/index/introducing-openai-o1-preview/
https://www.wired.com/story/openai-o1-strawberry-problem-reasoning/
If the price hikes released with Strawberry et al are what they are reported then hoping Open Source catches up real quick
I was having a conversation earlier in the week and I incorrectly stated that in the Netflix operating model SRE are responsible for reliability, this was incorrect. it is in fact the Critical Operations and Reliability Engineering team at Netflix (CORE) and is responsible for the reliability of the Netflix service as a whole. SRE is a component of this along with Applied Resilience Engineers, and Performance Engineers. Interestingly they do not follow the ITIL model
Meme that made me smile this week:
Interesting to see where Deloitte feel the biggest gains in financial services will be with generative AI, feels very generic and almost like it was written by AI. Not sure I agree with a lot of it.
The AWS CEO Matt Garman is completely on point but the light at the end of the tunnel is a mirage, the world class devs will adapt into architect roles but I have only met a handful of them in my career, its going to be a tough time for many. I have thought this for over 2 years from back when I was involved in early day building of coding tools at Globo-mega bank.
Grace Hopper is an absolute queen and you can watch one of her lectures here on protecting information in the future from 1982
https://arstechnica.com/science/2024/08/we-can-now-watch-grace-hoppers-famed-1982-lecture-on-youtube
Want to deploy an agent real fast, Anthropic Quickstarts has your back
https://github.com/anthropics/anthropic-quickstarts
Interesting paper showing how large language model agents are now capable of exceeding domain experts across meaningful tasks on scientific literature and this was before RF on the Inference aka Strawberry became a thing.
https://storage.googleapis.com/fh-public/paperqa/Language_Agents_Science.pdf
The Enterprise Intelligence space is on fire right now and is a lesson in Porter's Five Forces. This YC open source offering from a company called Quivr is a doozy. It has 11 integrations and a Choose Your Own Model capability for companies to turn enterprise knowledge into a personal AI assistant. Repo with 35k stars here and the marketing is around creating GenAI second brains which I had to look up but is a fancy way of describing an AI assistant that not only stores and retrieves information but also comprehends, analyses, and synthesizes it in ways that enhance your decision-making and problem-solving abilities.
AI and War is big business right now, if Google will not do it then companies like Helsing will
https://time.com/7013685/google-ai-deepmind-military-contracts-israel/?utm_source=tldrai
Anthropic has released workspaces that provide an abstraction layer for your overall organization and individual API keys
https://www.anthropic.com/news/workspaces
There is money to be made in black market jail broken LLMs by all accounts
https://www.fastcompany.com/91184474/black-market-ai-chatbots-thriving