Weekly-ish Link Roundup #17
Follow the money, build models, and gaming, really?
Here are the Finalists for the 59th Nebula Awards: I know everyone is busy reading excellent works like Ethan Mollick’s, “Co-Intelligence: Living and Working with AI” (just kidding, it’s not out till April 2), but I will always make an argument that reading scifi is important. Think of it as stretching before you run.
This Startup Cleans Up The "Ugly Underside" Of AI: This article is paywalled but the lede is right here - “Unstructured, which preps sloppy data for LLM training, has raised $40 million at a $230 million valuation.” - this is the key if you want to look for the companies and technologies that will power this accelerating curve of innovations.
LinkedIn plans to add gaming to its platform: OK…this one gets the FacePalm of the Week Award (and it’s only Tuesday). I’m all for companies trying new things but seriously LinkedIn….games? Not a feature that allows greater visibility in your amazing social graph or building a machine-readable standard for resumes so that all the resume-ingestion systems out there can accurately ingest our resumes. No, let’s work games and then rank companies against the scores of their employees. Yeah, that’s the way forward. “TechCrunch has learned and confirmed that LinkedIn is working on a new games experience. It will be doing so by tapping into the same wave of puzzle-mania that helped simple games like Wordle find viral success and millions of players. Three early efforts are games called “Queens”, “Inference” and “Crossclimb.”
App researchers have started to find code that points to the work LinkedIn is doing. One of them, Nima Owji, said that one idea LinkedIn appears to be experimenting with involves player scores being organised by places of work, with companies getting “ranked” by those scores.
Business leaders need to shift away from product roadmaps – here’s how: I think this is a good read and agree with a lot of it. The part that rings clearest for me is the tension between having a set roadmap that allows you to schedule resources and calendar feature or product launches. We can’t get rid of that in its entirety but we do need a system that allows for greater flexibility especially as we move into an era of ever faster change.
Google researchers unveil ‘VLOGGER’, an AI that can bring still photos to life: “Google researchers have developed a new artificial intelligence system that can generate lifelike videos of people speaking, gesturing and moving — from just a single still photo. The technology, called VLOGGER, relies on advanced machine learning models to synthesize startlingly realistic footage, opening up a range of potential applications while also raising concerns around deepfakes and misinformation.” > > In the words of Paul Virilio, “when you invent the ship, you invent the shipwreck.”
Nvidia Backs Little-Known Upstart in India’s Biggest AI Bet Yet: Another paywalled article but here is the key point for free “Yotta and its 52-year-old CEO are buying thousands of Nvidia chips to offer AI capabilities in the country.” Right now, whoever is buying the most Nvidia chips are the ones acquiring the biggest forward-facing technology war chest. It’s really easy to follow the money right now.
ASCII art elicits harmful responses from 5 major AI chatbots: This is one of those stories that I love and hate. I hate that people have figured out how to trick LLMs into actions that could cause harm to people. And “love” is probably too strong a word but there is something about watching humans continue the long tradition of hacking systems. If you want a great example, look up phreaking and the Cap’n Crunch whistles.
Hey YouTube creators, it’s time to start labeling AI-generated content in your videos: “Starting Monday, YouTube creators will be required to label when realistic-looking videos were made using artificial intelligence, part of a broader effort by the company to be transparent about content that could otherwise confuse or mislead users. When a user uploads a video to the site, they will see a checklist asking if their content makes a real person say or do something they didn’t do, alters footage of a real place or event, or depicts a realistic-looking scene that didn’t actually occur.” > > Here’s why this is a Signal to me - go the grocery store and look at the labels on things like eggs and milk. Try to figure out what the difference is between organic, free range and cage free. I think that AI will push us to develop systems of labeling for content that we can actually trust and that will be machine-readable.
Meet the Tech Company That Had a Better Year Than Nvidia: So one of the reasons why AI is having so much of an impact so quickly, is that unlike other tech revolutions like the PC, it didn’t require any additional end-user infrastructure to access. It worked on your browser - boom - just like that. It also didn’t require strictly new tech on the backend or the infrastructure side. That’s changing rapidly. Remember the adage that in a gold rush, look for the people selling picks and shovels. That’s where the long money is. Same here. We’re going to learn a new pantheon of companies that provide that ever more important infra side. Leaders will need to understand this environment to better evaluate the AI solutions that people will be trying to sell them.
Apple researchers achieve breakthroughs in multimodal AI as company ramps up investments: Whatever Apple does is an important Signal. Here’s what the headline means “Apple researchers have developed new methods for training large language models on both text and images, enabling more powerful and flexible AI systems, in what could be a significant advance for artificial intelligence and for future Apple products.”
Invoke rolls out Workflows, AI tools for game developers: If you don’t think this is already happening in almost every field, you’re just not paying attention. “Invoke Enterprise developed Workflows for large game studios (as well as the entertainment industry), which have large teams that include less tech-savvy users. Workflows’s features include control over training assets and brand-specific style, ease of deployment for “non-technical” team members and security of generated images. It’s also built, says Invoke, with game developers in mind rather than the casual user.”
A 4-Quadrant Approach to Introducing AI in Your Business: I like this approach - I think this is the perfect time to be trying out models and seeing how they do. The one thing I hope doesn’t happen here is that we get wedded too early to models in a time when things, even fundamental things, are changing so rapidly.
Who Profits the Most From Generative AI?: > > Follow the money :- )
Google joins collaborative efforts to build localized large language models: My concern here is specific to the use of LLMs with human language. While it would seem like human language is bound by rules - e.g. grammar - it is also wildly idiosyncratic and bound within cultural contexts. It’s not so much that I worry about AI getting translations wrong but rather flattening translations such that we lose cultural and regional differences.
—Music Break—
Watch an Incredible Performance of “Take Five” by the Dave Brubeck Quartet (1964): Probably my absolute fav jazz song. And how important it was/is: How Dave Brubeck’s Time Out Changed Jazz Music.