Your weekly update on how AI is changing our lives. Our experts keep it clear and simple, so you can stay ahead of the game. This week we are focussing on ChatGPT's new tricks. Don't forget to explore our Archive and Share & Subscribe with your friends!
Much like Christmas, ChatGPTās birthday seems to come around faster every year, which is an impressive feat given that itās still less than a year old. That didnāt stop one major newspaper printing a āOne Year of ChatGPT'ā retrospective last week - just shy of two months early! - but we wonāt be too critical. OpenAI isnāt waiting around for ChatGPTās anniversary either, and the deluge of new features released in the last week could make retrospectives tricky to write. Itās interesting timing because OpenAI has been fairly quiet since the last time they shook everyone up, with GPT-4ās launch back in March. Since then their rivals have caught up - and even overtaken - ChatGPT in various areas, with Anthropicās Claude offering longer context windows, Bing Chat searching the internet for you, and Google Bard flaunting its fancy fact checking button. Whatās more, everythingās multimodal now; imagine your chatbot being unable to read image files as if it were still Q2 of 2023!
Time for OpenAI to strike back, it seems. As of this month, the bot can now finally process images, but multimodality is just the beginning of its new tricks. ChatGPT is now also a fully fledged text-to-image service, with DALL-E 3 integrated directly into its prompting interface. This is revolutionary stuff, but also hard to spot; weāve generated all the images in this weekās newsletter with ChatGPT, and youād be hard pressed to tell them from the outputs of dedicated services. The process of creating them, however, is both new and fascinating; a back-and-forth collaborative design process with an LLM, offering an early glimpse at what work may be like for many more creatives in the near future.
Thereās big news in other areas, too, not least the return of the much-missed Browse With Bing feature, which allowed ChatGPT to search the web for up-to-date information. We first saw a beta version of this GPT-4 mode in June, but it proved to be so good at browsing that it began circumventing content paywalls, and was swiftly shut down. Itās now back - hopefully for good! - and should put paid to the botās plaintive cries of āas of my knowledge cutoff of January 2022,ā plugging ChatGPTās biggest weakness and cementing its place at the top of the chatbot pile, in plenty of time for its first birthday. Why not celebrate early with our in-depth guide to ChatGPT/DALL-E 3 prompting?
Beta Testers Wanted!!
SPYGAMES is the thrilling new experience where youāll jump, climb, throw and dodge in fun immersive challenges developed with CIA experts to stretch your physical and mental agility. ā Click the link below to get SPECIAL access as a beta tester.
Jon Stewart and AppleTV have parted ways over the editorial direction of his flagship show, The Problem With Jon Stewart, with reports that Apple were unhappy about planned episodes about China and AI.
A team of researchers have discovered that ChatGPT can infer personal details about users from seemingly anonymous text snippets - and even social media posts - with high levels of accuracy.
Everything OK with you two? Maybe machine learning can answer that; a University of Oklahoma study has compared six algorithms used to predict impending divorce, with a success rate topping 98%.
Charlie Brooker - showrunner of Black Mirror and expert on nightmarish dystopian futures - has weighed in on the LLM debate by claiming AIs are ānot messy enoughā to supplant humans in the creative process.
Pras, rapper of Ghetto Supastar and one third of The Fugees, has moved for a retrial of his recent conviction for criminal conspiracy, alleging that his lawyer used generative AI to write his closing comments.
At SPY HQ youāll explore hidden worlds, break codes, run surveillance and spot liars - while a system developed with MI6 experts reveals your personal spy role and profile.