AI Top-of-Mind for 3.26.24 — Apple and MMLMs
To-of-mind for today is Apple’s recent publication on MLLMs (Multi-model Large Language Models). Well worth reading on its training and architecture, with the conclusion here:
We study how to build performant MLLMs. Through carefully ablating modeling and data choices, we identify important lessons that yields a pre-trained model achieving SOTA results on a range of few-shot evaluations. After SFT, this model family produces competitive performance on a wide range of benchmarks, while enabling multi-image reasoning and few-shot prompting. We hope that the identified lessons will help the community in building strong models beyond the any single specific model architecture or data strategy.
The depth of thought for the model’s responses is incredible, as depicted below. We’ll hear more on this.
A few additional items on models. The first addresses biases, a topic central to any training. Cassie Kozyrokv looks at this, along with issues that could appear if going too far. Remember Google’s Gemini? Then Eivind Kjosbakken in ‘Towards Data Science’ dives into how to implement RAG (Retrieval-Augmented Generation). And then Amirarsalan Rajabi in the same publication describes building your own text and voice local LLM. From his posting:
In this tutorial we will create a personal local LLM assistant, that you can talk to. You will be able to record your voice using your microphone and send to the LLM. The LLM will return the answer with text AND speech.
SXSW (South by Southwest) wasn’t that long ago, and Elaine in ‘UX Collective’ looks at some AI and design insights. These include product design, the evolution of creativity, and trends to look for in 2024. A great diagram from the posting:
On the retail front, what happens when you rely on AI just too much and don’t offer credit where credit is due. Well, Under Armor is now paying the price, with commentary by ‘Tech Crunch.’ But AI-based content creation won’t go away, and ‘SmartBrief’ draws the following conclusions:
The future of content creation “might not look the same as it does now or be completely different, but there will always be a place for creative, persuasive minds,”
Lastly on retail, do malls have a ‘tone’? ‘Forbes’ looks at virtual influencers and the positive impact on foot traffic, especially amongst the younger generations.
