AI News of the week – 16th June: A quick summary

Posted on by AI Enthusiast
The Week in AI - 16th June

This past week, Video GenAI blew up. With the arrival of Kling (which we discussed last week) and of course Luma Labs opening up access to everyone, the internet is awash with examples of what is possible. A foretaste of the future, if you will. Stability AI released SD3 and being open source, also the model weights. There’s a new LLM benchmark in town: LiveBench AI. OpenAI has shares some sobering updates and some good news. And of course, a summary of AI at Apple’s WWDC. Let’s dive in:

Disrupting the advertising industry

Someone made a Mad Max Beer commercial in 1 hour using Kling.

Granted that it has a long way to go, if something like this were pitched as a proof of concept, it would probably get approved in an instant. So yes, we definitely think that AI is disrupting the film making industry. May not be in the most obvious of ways, but it’s coming.

Dream Machine by Luma Labs

After Sora from OpenAI, Dream Machine has been the a revelation as far as the future of video AI is concerned. A lot of the original videos that were done on Sora by some very gifted artists, still had some post processing. Dream Machine seems to have a level of finesses much beyond that.

There are some really mind blowing examples that are out in the wild right now. Check out “Break the Tie”:

While there are still many limitations as far as what is possible for professional use cases. This is just a taste of the future. Something like this makes conceptualization, ideation and even pitching ideas infinitely easier. We see this as a democratization of talent. Anyone with an idea can attempt to bring it to life. We think the immediate use cases will be more pre-production than actual live content. But that’s a great start too.

Stable Diffusion 3 Medium is now live

Stability AI’s most sophisticated Image Generation model is live along with its weights. The focus of this release has been better realism, high accuracy at text output and of course the fact that it’s more efficient.

You can read more about the release here.

LiveBench AI – A new LLM Benchmarking Platform

Abacus AI has sponsored the development and release of a new ‘Contamination free’ LLM benchmarking platform. LiveBench was built to bypass the ability to hack through the current benchmarks just for high scores. Developed in conjunction with Yann LeCun (and his team), LiveBench aims at providing an evaluation of LLMs on reasoning, coding, writing and data analysis. Fight ⚔️! More on LiveBench here.

LiveBench AI Benchmarking Tool

OpenAI Updates of the week

Somehow they manage to stay in the news, week after week. I’m convinced Sam Altman is the other side of the Elon Musk coin. Equally controversial, yet much more palatable to the average netizen.

This week Mira Murati, CTO – OpenAI shared that none of their internal models are that far ahead of what is publicly available. Much in line with their philosophy to do regular releases to ensure that AI advancements seem smooth, rather than like a bat to the face. As was the case when ChatGPT went from 3.5 to 4. Either they are setting the world up for another huge upset (which would be very much like what one has come to expect from Sam Altman.) Or AI development is slowing down? 🤷‍♂️

In other OpenAI news, Elon Musk has dropped his suit against Sam Altman and OpenAI. No one knows why. And OpenAI has added a retired head of the NSA as one of their board members. This one is a real 5-D Chess move if you think about it. The (ex)head of the largest spy agency on the planet is on the board of the company that is touted to birth AGI, which will have more data and privacy implications than anything else in the history of modern man. 🤯 Of course, he’s part of the Safety and Security Committee.

AI = Apple Intelligence

MKBHD has pointed out in the past the great lengths that Apple has gone to, just to avoid using the term AI. This was the first time Apple broke that rule. We already knew that Apple would not be looking at AI the way everyone else does. They were, as per usual, late to the party. But they do make quite an entrance.

Ignoring all the fancy GenAI text-to-image and text-to-video applications, Apple chose instead work GenAI into their apps at a more integrated level. Siri is much more useful. We definitely don’t see any reason to complain when you see the magic accomplished with the Apple Tablet: (see below)

Math Notes in action – 5 mins should give you an idea

Apple Intelligence in 5 mins:

Integrated AI in the most practical way possible

Privacy and other concerns were taken care of in typical Apple style. The TL;DR is that you don’t need to worry. Apple will do everything in its power to keep your data safe. And since Apple is doing AI differently, they also announced a partnership with OpenAI to use ChatGPT as an additional layer of support in situations where it isn’t possible for Apple to solve your query. This is also done after getting explicit consent from the user. Apple has also stated that even they can’t access user data, so as far as privacy is concerned, they have left no stone unturned. This partnership, of course, pissed off Elon Musk who has threatened to ban Apple devices at his businesses. We are not qualified enough to state whether his concerns are founded, but get the sense that it might be an overreaction. The real head scratcher for us is why would Microsoft be subsidizing OpenAI for Apple to use?

Anyway, that’s the wrap for the week. If you found this useful, please share this post with anyone who might want to catch-up on the world of AI.

See you in the next one!

Related Posts

Leave a Reply

Your email address will not be published. Required fields are marked *

WhatsApp
Follow by Email
Facebook
X (Twitter)
LinkedIn
LinkedIn
Share