In these screenshots our (illustrative, toy) task is to write a one-line description of a cleaning product. A base LLM will not respect instructions to this effect. The task must be communicated by repeating k-shot examples, which we generate using Claude.
LLMS
-
Every AI Model Since GPT-4 Looks Identical
By
–
Update: On this graph, literally every model that has come out since GPT-4 still looks pretty much like … GPT-4.
-
LLMs for Coding: Productivity and Educational Value Considerations
By
–
I'm not against LLMs for coding. I think there's a range of situations where they save time and have educational value. However, in my own workflows I have not found them to lead to any productivity boost. Mind you, I was also virtually never using SO in the pre-LLM days, whereas
-
LLM Limitations in Coding: Expert Skepticism Grows
By
–
Coding used to seem like the best use for LLMs, but in recent days @burkov and @fchollet have both come out strongly against even that use case. Both tried; both have been left disappointed.
-
5 Best AI Chatbot Apps for Android Reviewed
By
–
My 5 favorite AI chatbot apps for Android – see what you can do with them: Beyond the official ChatGPT app for Android, there are several helpful and interesting alternatives. These are the five chatbot apps I reach for most… https://
zdnet.com/article/my-5-f
avorite-ai-chatbot-apps-for-android-see-what-you-can-do-with-them/?utm_source=dlvr.it&utm_medium=twitter#ftag=RSSbaffb68
… #BigData #DataScience #IoT -
Clarifying GPT-3’s WebText training data
By
–
This is historically false. The WebText dataset used (among others) to train GPT-3 consists of the scraped content of URLs linked from Reddit posts with at least 3 karma, including downvotes. You don’t scrape URLs that only appear in low-karma posts because they’re mostly spam.
-
Large-Scale ML Training Systems and Language Model CO2e Emissions
By
–
I gave a talk at the MLSys conference in May this year, touching on various topics, including large-scale ML training systems, abstractions for embedding ML choices in computer systems, and CO2e emissions of language model training. https://
mlsys.org/virtual/2024/i
nvited-talk/2592
… -
GPT-4 base is not aligned without SFT or RLHF
By
–
That’s why I qualified “aligned” — GPT-4 base isn’t aligned (i.e. no instruct SFT or RLHF).
-
Clarifying what “aligned” means in AI industry
By
–
No, that is what it means. Philosophers of AI safety often mean something bigger, but in industry “aligned” just means “tuned for use” and implies some combination of SFT or RL.
-
Webinar Recording: Agentic Architecture with Toolhouse AI and Groq
By
–
If you missed our last webinar with @ToolhouseAI
, watch the recording to get a high level overview of agentic architecture, the current challenges of building agents, and how you can overcome them with Toolhouse AI and Groq. https://
hubs.la/Q02L-rz40