Happy Holid-AI-s

It's less than two weeks until Christmas — heads up if you still have shopping to do!

Between that, Hanukkah, Kwanzaa, New Year's, the Winter Solstice, and the general air of festivity around this dark and dreary — yet, at least in snowy climates, starkly beautiful — time of year, it's of course a great time to connect with friends and loved ones.

But according to some of the leading AI model making companies, it's also a great time to launch powerful new products, services, and features.

If you've been less tuned into tech news than usual due to the impending holidays and associated obligations, you can be forgiven for missing the news. Here's a quick rundown to keep you updated:

OpenAI is in the middle of a wave of consecutive holiday-themed announcements fittingly titled the "12 Days of OpenAI" in reference to the Christmas tradition. The biggest among the news is arguably the public launch of Sora, its long-gestating AI video generation model that was shown off back in February 2024.

Early reactions to Sora's arrival from AI power users and filmmakers have been mixed, but my initial tests have been largely impressive, if lacking some of the finer detail and high quality visuals of rival Runway 's Gen-3 Alpha model or Luma AI 's Dream Machine.

And yet, OpenAI wasn't the only company to launch a new video model just before the end of the year. The oft overlooked Pika came out swinging as well today with the launch of its new Pika 2.0 AI video generator that lets users upload their own custom assets — such as characters, props, settings, and more —?and then automatically inputs these into AI generated videos smoothly and easily. This is an industry first from what I'm aware, and makes Pika much more useful as a possible tool for creating brand advertisements.

Furthermore, 谷歌 debuted its own laundry list of new AI updates including experimental programs like Project Mariner that lets the company's brand new Gemini 2 underlying large multimodal model actually control your web browsing experience and click on things for you. However, it's only available on a waitlist access for now.

VentureBeat founder Matt Marshall just made the case that Gemini 2 Flash and its multimodal live streaming capabilities — allowing the user to screenshare and speak with Gemini 2 about the content onscreen, asking it for tips on how to edit videos or perform certain tasks — is tantamount to an iPhone-launch level of importance.

Meanwhile, VentureBeat's executive editor Michael Nu?ez reported on how 微软 wasn't about to be left out of the holiday-themed AI fun, launching Phi-4, the latest and most powerful and performant version of its smaller language models, carving its own path separate and distinct from its big investment in OpenAI.

And as Senior AI Reporter Emilia David reported, OpenAI's new ChatGPT Projects feature (announced today), launch of ChatGPT's own screensharing and video analysis capabilities, and its expansion of the ChatGPT Canvas feature continue to augment and supercharge the most popular product of the generative AI era, making the new $200 monthly Pro tier perhaps more enticing, even to those who are feeling a cash crunch from all their other holiday spending.

(The launch of a new interactive Santa voice in ChatGPT — which developers can also plug into their apps — also was seasonally appropriate and a great touch.)

While I've heard speculation the rush of new AI announcements is driven at least in some part by a desire among engineering teams within the tech companies to secure milestones that will unlock year-end bonuses, I'm still gobsmacked by the amount of new stuff shipping in the last few weeks of 2024, and I think there's more to it than simple dollar eyes.

No, it seems as though even 2 years and counting into the generative AI craze started by the launch of ChatGPT, there's still lots of advancement and improvement happening rapidly and concurrently across the AI industry. To my eyes, a lot of the groundbreaking research and tech advancement that has occurred inside labs is finally being "productized," or put to work in applications and services that, hopefully for all of us, make our work easier, better, faster, more efficient and more enjoyable — and, best of all, more expansive and ambitious, allowing us to create things never before possible on our budgets or with prior tech.

All of which is to say, I kind of feel like the proverbial "kid on Christmas morning" with all these big new AI product launches, and will likely spend some of my holiday break playing around with them as I would have a microscope or erector set in my youth.

That's all for this week. Thanks for reading, writing, subscribing, sharing, liking, commenting, and being you.

— Carl Franzen

Read More