AI Top-of-Mind for 4.18.24 — Election Avatars
Top-of-mind is election season. As India goes to the polls, voters there are experiencing the use of AI a generation beyond previous years, with avatars addressing voters by name and in their native languages. The ‘NY Times’ dives into details and the technology. Coming soon to an election near you! Scared yet?
On the same line of ‘what-if,’ the ‘EFF’ published a good essay on the perceived tech apocalypse, with a conclusion that today’s AI is neither artificial or intelligent. But taking it one step further, if AI is so disruptive and powerful, should it be nationalized like the bomb? Anthony Lawrence in ‘Tech and Me’ concludes ‘no,’ but I’m sure others are asking the question. And are we smart enough to control the genie? Daniel Sexton in ‘DataDrivenInvestor’ looks at how we’ve adapted over the decades, and in fact have, by all accounts, become more intelligent. And his takeaway:
The next human intelligence innovation will be an explosion of creativity and intuition.
Moving to models, or at least the companies developing them, more on Mistral AI. ‘The Information’ reportingon a planned round at a $5B valuation, a pretty good multiple on $22M in revenues. Also from the article:
· Mistral has also admitted to using Meta’s open-source AI, Llama 2, to create its own AI, only disclosing that fact after the information leaked.
· In February, Microsoft took a tiny minority stake in Mistral. Microsoft said it would offer Mistral’s AI models, including its most advanced one, Mistral Large, on its Azure cloud rental service. Mistral Large was designed to compete with OpenAI’s most advanced model, GPT-4.
Further afield, Ignacio de Gregorio dives into Google’s ‘Mixture-of-Depths’ architecture, capable of dynamically allocating compute to each prediction. A link to the original DeepMind paper. And if you are into understanding layers, have we gone overboard on depth? Salvatore Raieli in ‘Level Up Coding’ concludes ‘yes,’ and how to readjust.
Then to the positives and negatives of creative. First is using AI to create business logos, and I admit I’ve dabbled with this. Jeff Hayward in ‘Ai-Ai-OH’ concludes we’re not yet there, and quotes graphic designer Danielle Hitchcock:
“So, can Ai generate a logo for you? Sure. Will you like it? Maybe for a while,” explains Hitchcock. “But can it create a successful brand? Nope. Absolutely not. Not yet. Hopefully, for me, not ever.”
Second is how to properly create hand-drawn diagrams with AI tools. Some good approaches, but there is a learning curve. More from Xinran Ma in ‘Bootcamp.’
And then the impact of AI on music. Jim Clyde Monge in ‘Generative AI’ looks at the state of the art, and the conclusion is that it is still too early to draw conclusions. But remember that the industry has survived other upheavals.
Related to creative, in a way, is the use of Gen AI by students. Another update by Enrique Dans on how we can embrace the proper use of AI. Includes a good background on ‘Turnitin’ if you are not familiar with the tool.
Lastly, a few times I’ve mentioned the gadgets announced at CES, including the Humane AI Pin. If you’ve had any thoughts about purchasing the thing, please watch Marques Brownlee’s ‘Youtube’ review, and the title — ‘The Worst Product I’ve Ever Reviewed… For Now’ — says it all! Sticking to videos, Boston Dynamics is retiring ‘Atlas’ after almost a decade. A great sendoff, and they are also teasing the next generation version, below on the right.
