Tweet
Starter post for AI related learning... Lets use this as a space to start posting AI related training topics
-
-
How I'd Learn AI in 2024 (if I could start over)
-
Here are some classes from MIT that are made available for FREE online. There are 9 Lectures available at this moment
Comment
-
AI itself is a very broad field, its almost like saying math. Some is more theoretical in nature and some is completely productionized and being used to practically solve problems today. Lately the buzz has been around Large Language models that model language as probability mass functions to be estimated by an LLM conditioned on earlier parts of the text being generated that tend to work as reasoning agents, unlocking use cases on what was previously thought to be only doable by a human. For a pertinent example asking an LLM to review batch records for discrepancies.
Because of how productionized these reasoning agents are today and relatively cheap (sub penny per request) its a very low bar of entry to start prototyping with this and start unlocking disruptive use cases. For example you can go to claude: https://claude.ai/login?returnTo=%2F%3F. And read the documentation and start seeing in action. Find public documentation (your companies legal team may not like it if you use proprietary data into an unvetted system). and throw it into the LLM and start asking questions about equpiment for example, a batch for example.
From there on, its a classic software engineering problem of wiring text to the LLM API calls and immediately you are making a huge impact. There are services for other kinds of AI too, like generating images, or classifying them, and documentation for how to set it up (again usually a software engineering endeavour).
This is separate from learning AI from scratch, the mathematical underpinning of how the models are trained, why the work and how to optimize them at a fundamental level, that takes statistics, information theory, calculus, optimization, linear algebra, vector calculus and then surveying Machine Learning in general (data science), deep learning. Then the road roughly depends on the domain, is this visual data we are working with or text or audio or perhaps all of them (multi modal) and the latest models and breakthroughs are rarely well documented in some book, because of how fast things are moving - you will have to read papers as the field evolves.Comment
-
Comment