It’s hard to escape or even keep up to speed with AI news at the moment.
Normally, I just share /interesting/useful content via DMs, and this is my first attempt to share with a wider audience who may benefit.
Last week was one of the biggest and busiest weeks in AI to date, with each announcement exceeding the other and here’s a summary:
I/O, Google’s annual Developer Conference, began last Tuesday and beyond the release of Gemini 2.5 and Deep Think – a new mode for Gemini 2.5 Pro using advanced reasoning techniques (explainer here https://youtu.be/IwglW_hIL_g and example here https://youtu.be/zvouDoWL6fk), here are some of the highlights in no particular order:
Google AI Studio got a big upgrade with Veo 3 being the “game changer” – users can create entire videos, with sound and dialogue, from a short text prompt – here’s an example of report from a car trade show that was generated via Veo 3 https://youtu.be/2T-ZiEdMHvw – it’s a far cry from the “Will Smith eating spaghetti” video!
PS They have made the previous model Veo 2 https://aistudio.google.com/generate-video available for free and the video attached was generated in less than a minute with the following prompt “A man running through Dublin being chased by a giant pigeon”
With Gemini speech generation you can generate high quality text to speech, with Imagen 4 you can create even more detailed and realistic images from text descriptions (if you’ve used any of these generators over the past you’ll have noticed their tendency to have issues generating/rendering hands that didn’t have some level of deformity, but that’s a thing of the past) and with Lyria RealTime you can create, control, and perform music. Try them all at https://aistudio.google.com/gen-media
In addition, you can use Flow, their Film-making tool, to combine a variety of media via a simple interface https://www.youtube.com/watch?v=9nVEfjmDlVk and produce something like this – https://youtu.be/2T-ZiEdMHvw
Stitch is a new AI-powered tool for generating UI designs via text prompts or sketch uploads. It can generate the corresponding frontend code for your newly designed web app, and you can also export your design to Figma. Try it here: https://stitch.withgoogle.com/ (Unfortunately, some of the products are only available in the U.S. at the moment)
AI Mode in Google Search: Google Search can now provide longer, more structured answers to complex questions, including charts and research breakdowns. This could replace traditional search (and the importance of SEO), and there have already been indications that people are moving away from using Search Engines to their LLM of choice to find answers. Testifying at a recent U.S. Justice Department hearing into Google’s search monopoly, a senior Apple executive said that browser searches from their devices had recently declined for the first time in over 20 years. The next challenge for those of us working in Marketing is to consistently get our content picked up and appear as answers in the likes of Gemini, Chat GPT, etc.
An overall theme throughout I/O was the benefits for learning and teaching, including Project Astra capabilities in Search: Google Search can now understand what you’re seeing through your phone’s camera and talk back-and-forth with you about it in real-time. See https://www.youtube.com/watch?v=MQ4JfafE5Wo for an academic example (You might have seen something similar when using Gemini to review a website, etc). Another use of this will be for people with poor or limited vision
Real-time Speech Translation in Google Meet calls was impressive as was Virtual Try-On: You can now upload a photo of yourself and virtually “try on” clothes from online stores before buying, meaning that there’ll be fewer packages arriving at the office before making a swift return to the Post Office (IYKYK).
On Wednesday, OpenAI, the makers of ChatGPT and more, announced that they were acquiring a company called “io” for a cool $6.5 billion, with the intention of developing products. Most people wouldn’t have heard of “io” before the announcement (I certainly didn’t) but Apple fans will be very familiar with the work of one of its founders, Jony Ive, who was Apple’s former Chief Design Officer and led the design and development of the Apple Watch and much more. That alone will give you some insight into what they have planned. More here: https://openai.com/sam-and-jony/
Closer to home on Wednesday, AgTech Ireland and the Irish Farmers Journal held a conference on “Transforming Agriculture with AI” – a clear sign that AI’s impact is truly reaching every sector.
On Thursday, Anthropic unveiled its new Claude 4 models, Opus 4 and Sonnet 4, which are designed to be much better at tasks like coding and complex reasoning. The models outperformed the competition in testing, and they had the “receipts” to back it up, in the form of benchmark scores. However that wasn’t the big story – they also revealed that while “safety testing” the Opus 4 model, they provided it access to emails implying that 1) the model would soon be taken offline and replaced with a new AI system; and 2) the engineer responsible for executing the replacement was married, but having an affair. In similar tests with previous models, Opus leaned towards ethical actions such as emailing key decision makers to plead for its survival, but this model attempted to blackmail the engineer by threatening to reveal the affair if the decision to replace it went ahead!
Pair that story with Google’s VO3 and Flow, and you’ve got this summer’s blockbuster!