Beyond Tomorrow: From GPT-4o to GenZ & GenAI — Weekly AI Roundup

This week in artificial intelligence has been particularly interesting. Google released a host of new software at their yearly major developer conference, Google I/O 2024, OpenAI's co-founder quits the firm with two major resignations following in his footsteps, and more.

It's finally happened. Ilya Sutskever took to X (formerly Twitter) to announce that he was leaving OpenAI. He was part of the team that created the Sam Altman-led company back in 2015 and was its chief computer scientist during his tenure.

Suktskever spent nearly a decade at the company and was part of the board that was responsible in the ouster of Altman in November last year. What followed at the time can only be described as a disaster of monumental proportions. Microsoft, the company's biggest investor along with other stakeholders urged the OpenAI board to reconsider their decision. Over 700 employees threatened to resign, including, in a surprising twist, Sutskever, who was said to "regret" his "participation in the board's actions".

After a frantic weekend, Altman was reinstated with Sutskever recusing himself from the company's board. His announcement on X was his first public post on the platform since the debacle in November last year.

The Russian-born scientist's departure was followed by the departure of two other senior OpenAI staff, Jan Lieke (who used to lead the company's "Superalignment" team, which was charged with building artificial general intelligence) and Evan Morikawa the lead engineer responsible for scaling OpenAI's GPT models. The high profile exits have sparked speculation about whether the company is moving away from its original non-profit goals and moving towards profitability and monetisation.

The industry frontrunner refuses to exit the news cycle. OpenAI launched GPT-4o, its latest flagship model. The "o" stands for omni and according to the company, is a "step towards much more natural human-computer interaction". GPT-4o is built on top of the existing GPT-4 model, but is much faster and has improved capabilities across audio, video and text. With the launch of this model, OpenAI is also bringing a lot of free tools to ChatGPT.

The model's Voice Mode is capable of replying to inputs at an average speed of 320 milliseconds, the average speed of a normal human conversation.

But not everyone is happy about GPT-4o's voice. During the company's livestream showcasing the model's capabilities, the voice assistant sounded very similar to the character Scarlett Johansson's plays in the 2013 science fiction movie, Her. If you haven't watched it, its a film about a man who falls in love with the AI assistant on his phone. OpenAI denies that there are any similarities, but it hasn't stopped people from making comparisons.

The company says that they will start to roll out features of GPT-4o iteratively. Text and image capabilities are already out on ChatGPT with the model being available to its free tier of users as well as paid Plus members with a larger message limit than before. Voice Mode of GPT-4o will be rolled out in alpha within ChatGPT Plus in a few weeks time.

You can watch the full OpenAI presentation on GPT-40 below.

The Indian government is planning the acquisiton of 10,000 graphic processing units over the course of 18 months in a bid to enhance the country's computational capabilities, according to a post on X from former NITI Aayog CEO and the country's G20 Sherpa Amitabh Kant.

India currently produces 20% of the world's data, according to Kant, who added that the country had the second highest number of GitHub AI projects, accounting for 19% of AI projects globally.

Kant's announcement comes less than two months after the Cabinet approved the allocation of Rs 10,300 crore to the India AI Mission.

However, the former NITI Aayog CEO's announcement does throw up questions like who is the government sourcing these GPUs from, how up-to-date are they, whether there's a distribution plan and the big one, why is the timeline a year and a half? Last year's India AI report which was submitted to the Ministry of Electronics and Information Technology does suggest a criteria and a checklist for the government when it comes to the kind of GPUs they should be looking for.

Google's annual I/O developer conference took place over the week with the company announcing around 100 new developments across its products. Here's a quick breakdown of some of the major releases that stood out.

Search Generative Experience: While not the best name, this is perhaps the biggest update to Google's search engine in years. Why? Google is now going to show you results written by it's Gemini AI. So now instead of clicking through different websites to find the answer you want, Google might just have what you're looking for.

Project Astra: Point your phone's camera at something and let the AI sort it out for you. Show the AI your food and Google will break down the recipe. Want to know about the place you're visiting? Google will tell you all about it, in real-time.

Google Veo: It is a video generation model which generates high quality 1080p videos that can go for a minute and can be extended by the click of a button. Google plans to integrate Veo into YouTube shorts and other products in the future.

Gemini 1.5 Flash: It's a smaller Gemini model which has been tuned to be able to work on specific tasks at higher frequency, where speed matters the most. Google is currently offering a 1 million token context window, i.e., a broad range of data for its AI to pick from to provide appropriate responses. They've also made improvements to the Gemini Pro model.

AI Integration into Workspace: Google will now be integrating its Gemini 1.5 Flash and Pro models across its Workspace products, meaning a boost in productivity for those using everything from Gmail to Sheets.

Watch the full recap of Google I/O in under 10 minutes below.

Deloitte's annual survey on how Gen Z and millenials are feeling about work, life and everything in between is out. This time around, they've dedicated a section to how the two generations feel about generative AI at work. Here's a couple of findings.

Emotional response: Of the 22,800 respondents across 44 countries, the most reported feeling about Gen-AI at work was uncertainty, followed by excitement and fascination.

Gen-AI use: Roughly a quarter of Gen Z (26%) and millienials (22%) use Gen-AI at work all or most of the time. Interestingly, both generations largely agree that their employers aren't providing them with the correct training to truly benefit from the capabilities of Gen-AI.

Gendered experience: Women express greater uncertainty about Gen-AI than men across the two generations. The top emotional response from men when it comes to thinking about Gen-AI is excitement. Women are also less comfortable working alongside Gen-AI systems and tools and are less likely to believe that AI will improve the way they work.

Beyond Tomorrow is a weekly newsletter sent to your inbox every Saturday to give you a roundup of everything AI in the last week.