Welcome to our roundup of this week's news on biologically engineered AI.
This week, AI took over Apple products and made us jealous of the iPhone 15.
A Chinese text-to-video generator could beat Sora.
And Google's AI search is ploughing through our electricity and water resources.
Let's start.
Search for resources
Google's Search Engine Experience (SGE) had a rocky start, but despite some strange suggestions, the AI Overviews feature is definitely quite useful. And extremely resource-hungry.
Guess how far more electricity and water a search with AI overviews uses than a standard search query. Guess again.
After reading this, you should have a special perspective on what it takes to search out that restaurant online or create a cute cat picture.
Apple Intelligence
Apple was slow to get off the starting blocks within the AI race, nevertheless it made up for lost time this week. At its WWDC conference, the corporate announced that latest Apple products would get a complete host of AI features, including an improved Siri.
Apple has announced that it has partnered with OpenAI to integrate access to ChatGPT from its apps, including Siri. Would you be OK with OpenAI accessing your emails, calls, and messages?
Elon Musk has made it pretty clear what he thinks would occur to your data.
pic.twitter.com/7OgZAAdPf6
Musk is a co-founder of OpenAI, but is unhappy with the direction the corporate has taken. In February, he filed a lawsuit claiming that OpenAI's pursuit of profit and secrecy of its models constituted a breach of contract.
This week, Musk withdrew his lawsuit against OpenAI Speculation is growing in regards to the reasons for this move. Jealousy of success? Or one lawsuit too many in light of the increasing allegations of insider trading?
No lights, no camera, Chinese motion
If your job involves operating a camera or organising a movie set, you're probably not excited in regards to the rapid pace of development in text-to-video.
We're still waiting for OpenAI to release Sora, but a Chinese company just released a text-to-video generator called Kling that appears really good. It generates clips twice so long as Sora at 1080p and 30fps.
The beta version of Kling, while OpenAI remains to be holding off on the discharge of Sora, offers an interesting insight into the East-West approach to AI “security.”
While we wait for Sora, Luma AI has just made its Dream Machine publicly available. It's free, so naturally the server is totally overloaded. It only generates 5-second clips, however the demos still look good. Expect a number of animated memes.
Introducing Dream Machine – a next-generation video model for creating high-quality, realistic recordings from text instructions and pictures using AI. It's available to everyone starting today! Try it without spending a dime here https://t.co/rBVWU50kTc #LumaDreamMachine pic.twitter.com/Ypmacd8E9z
I HAVE
Meta plans to launch its Meta AI functionality in Europe, but the corporate says it needs somewhat help doing so. It desires to train its AI using social media posts from European users, but there’s major resistance from data protection watchdogs.
Would you permit Meta to make use of your Facebook and Instagram content to coach its AI? Stupid query. If you're not within the EU, then Meta has been doing this for a while without asking your permission.
If all the EU's younger, tech-savvy social media users drop out and only the training data of the older ones stays, one has to wonder how Llama will evolve.
Another AI company that has to cope with GDPR rules and EU AI law is the French startup Mistral AI.
The company just secured $640 million in its second round of funding and its AI models are difficult big players like OpenAI and Google.
As AI applications evolve, we’d like latest benchmarks to assist us make meaningful comparisons.
The next step in AI evolution is to make use of AI as an agent to perform your on a regular basis tasks across platforms. Google DeepMind has developed NATURAL PLAN, a benchmark that measures LLMs using natural language planning.
Which LLM is best for planning a visit or arranging a gathering with friends? It seems that all of them do a reasonably bad job of it. And GPT-4o is worse than GPT-3.5 in some cases.
AI Events
Two interesting AI events took place this week:
At the AI Summit London 2024, speakers from OpenAI and NASA discussed the newest developments in AI and real-world applications of this transformative technology.
The AI Accelerator Institute hosted the Generative AI Summit New York 2024, bringing together a vibrant community pushing the boundaries of what’s possible with generative AI.
If you might be in Canada next week and need to see how AI is transforming the general public sector, take a look at the CDAO Canada Public Sector event.
In other news…
Here are another click-worthy AI stories we enjoyed this week:
And this can be a wrap.
I have a look at my iPhone and wonder if an upgrade is warranted to check out Apple's latest AI features. Do you think that Elon Musk is correct to be concerned about OpenAI getting hold of user data?
If you reside in Europe, we would really like to know for those who are for or against Meta training its AI together with your social media posts. If you reside outside the EU, would you choose out if possible?
Perhaps the ideas of being connected to actually useful AI devices and preserving privacy are mutually exclusive. The only solution appears to be on-device AI.
Let us know what you think that, chat with us on X, and keep sending us links to AI news.
