Another paper pointing out in details what we've known for a while: LLMs (used via prompting) cannot make sense of situations that substantially differ from the situations found in their training data. Which is to say, LLMs do not possess general intelligence to any meaningful degree.
What LLMs can be good for, is to serve as knowledge/routine stores for an actual AGI. They're a memory -- a representation of a data corpus -- and memory is a necessary component of intelligence. But keep in mind that intelligence is not just memory.
🎊 Go 1.22.4 and 1.21.11 are released!
🔏 Security: Includes security fixes for archive/zip and net/netip
🗣 Announcement: https://t.co/iPV5aKjfD0
⬇️ Download: https://t.co/GHe3QnGFAp
#golang
Gemini 1.5 Flash is one of my go-to models for scaling use cases that need millions of monthly calls. It helps me strike a good balance between quality, speed, reasoning, and ROI.
Big news – Gemini 1.5 Flash, Pro and Advanced results are out!🔥
- Gemini 1.5 Pro/Advanced at #2, closing in on GPT-4o
- Gemini 1.5 Flash at #9, outperforming Llama-3-70b and nearly reaching GPT-4-0125 (!)
Pro is significantly stronger than its April version. Flash’s cost, capabilities, and unmatched context length make it a market game-changer!
Huge congrats to @GoogleDeepMind on the incredible
Gemini launches! Can't wait to see what new applications Gemini unlocks!
More breakdown analysis below👇
Good news for @Google Developers:
- Gemini 1.5 Flash and 1.5 Pro are now GA
- Gemini 1.5 Flash now has a 1,000 RPM limit
- Gemini 1.5 Flash tuning announced
- JSON Schema mode is now available in the API
- AI Studio Mobile support + light mode are live
https://t.co/vkOgeXhMH4
🦜🔗LangChain Release Notes Week of 12/11
🎯LangChain Core, LangChain Community, and LangChain v0.1
💬New LLM integrations like Mixtral and Gemini
📃5 New LangChain Templates
🧑🤝🧑Lot's of community apps to highlight
Read here & sign up to get it in your inbox every other week: https://t.co/BkEZpIk8ow
I'm *thrilled* that my support for @GoogleCloudTech Vertex AI Matching Engine has finally landed in @langchain JS.
This means you can store your docs with GCS, embed your docs with Vertex AI Palm API, and index them with Matching Engine all on one platform!
We’re updating Bard with a new technique called implicit code execution. Now it runs code in the background when it detects computational prompts, improving the accuracy of word and math problems by ~30%.
https://t.co/yHayR1VhvT
@GoogleCloudTech What fascinates me the most is the large set of use cases popping up daily. Many of us are figuring out how to create and capture value with responsible usage of GenAI. Looking forward to connecting with more folks working in this field.
What are Large Language Models (LLMs)? https://t.co/SyX5EipQIp < if you've been nodding along with all this AI talk but secretly don't get it, you're not alone. Most of us are faking it. But this video with Dale and Nikita does an EXCELLENT job at explaining the basics of LLMs.
The #GoogleCloud Innovators community is buzzing with excitement about #GenerativeAI! Connecting with fellow tech enthusiasts is an excellent resource for discovering fresh ideas.
Let's keep the conversation going!
https://t.co/ALc0y1wN6B
Does your team use "architecture design records" or something similar to capture design choices your team made?
We just published a terrific short @googlecloud doc that describes when to use ADRs, how to format them, and how they work.
https://t.co/RS2hYZwarh
Yup, @Google Bard does a pretty great job at coding tasks now.
Code generation, code explanation (try pointing to a repo!), fixing code bugs, and even giving a source citation where applicable.
https://t.co/pcH0SZiCU9
I even got COBOL back:
Excellent explanation of Spanner's consistency guarantees: it can run on thousands of nodes, but to applications it looks like a single, consistent database. Sounds like magic but after reading this article you'll know how it works :-)
@googlecloud https://t.co/m5cm352QdS
Hope you’ve had a chance to start testing out Bard! We’re constantly working on developing new features for it - today I wanted to share that we’ve improved Bard’s capabilities in math and logic by incorporating some of the advances we’ve developed in PaLM. /1
Today we announced a number of new innovations and partnerships across our data analytics and AI portfolio to help our customers make the most of their cloud investments and deliver new experiences for their customers. https://t.co/BIH9yuoTZ1
If you're building event-driven systems, don't start from scratch. Use Cloud Code templates to jumpstart your code, Eventarc to connect to event sources, and Workflows to orchestrate it all, as @rseroter shows in his latest blog post and demo
Ok, so @google Bard is now opened up to more folks (https://t.co/5khJoT7XEF).
I just asked it the most controversial question I could think of: "Is a hot dog a sandwich?" Well done, Bard. Well done.
Excited about PaLM API: an easy and safe way for developers to build on top of our language models, and MakerSuite, a tool to jumpstart prototyping - both in private preview today. @googlecloud customers can also access these models + more via Vertex AI.
https://t.co/ajijHlgdMh
The best way to achieve agility is to trust the teams. Connect them with customers and let them collaborate on what to do next. Build small. get feedback. Adjust.
Forcing some Agile™ process on the teams is the opposite of that. No trust in sight. Lots of control and bullying.