AI Top-of-Mind for Jan 17
Top-of-mind are two great LLM tutorials for beginners, the first, titled ‘A jargon-free explanation of how AI large language models work’ in ‘Ars Technica,’ walks you through all of the basic concepts.
The second by John Loewen in ‘DataDrivenInvestor’ is a cookbook to creating better maps and charts with GPT-4. Speaking of LLMs, a technique to offload components of Mixtral-8x7B to permit running on your typical high-end PC. Benjamin Marie in ‘Towards Data Science’ details the solution, and a link to the source code is here.
On the corporate front, if you didn’t catch it saturating the news yesterday, Musk is ‘requesting’ to grow his Tesla holding to 25% if he is to make additional AI investments in the company. And, looking back to the IMF posting yesterday, one additional proof point, this from Duolingo as they replace translation contractors with AI. More from Brian Kean in ‘Medium.’
Moving to creative, techniques to generate convincing optical illusions with AI image tools. More from Miroslav Novohradsky in ‘Generative AI.’
And on the not so positive creative front, a view by James Presbitero in ‘Practice in Public on how to identify text created by AI. I’m sure we all have our own takes on what words or phrases are dead giveaways. Quite transformative. Just kidding!
Lastly, on retail, one survey by ‘MessageGears’ looking at AI as helpful in determining purchase intent. Some observations include AI delivering significant or moderate savings in 72% of cases, and 53% state that AI is helpful in determining which customers are likely to purchase. More in the ‘MediaPost’ article. And then Salesforce has announced new AI-driven tools for chat personalization, commerce site design, and target segment creation. Details are at ‘MarTech.’