☕️ Bard Gets a Taste of Gemini

AI Tangle Newsletter

The world of tech and AI gave us a peek at Google Maps’ next-gen search experience and a new version of Bard, now powered by Google’s Gemini Pro model. Meanwhile, the French at startup Twin Labs is looking to put vision-based models to good use, and UMG makes a bold statement against the likes of Chinese social media giant TikTok. Join us at AI Tangle as we untangle the world of AI.

Google Bard's next significant upgrade is its integration with Gemini Pro​

Announced on Thursday, Google has unveiled the next iteration of Bard, which will now leverage the company's latest and most capable generative AI model Gemini, with support for more than 40 languages, additionally adding Imagen 2 for image generation. Google stated that the chatbot will improve understanding, summarizing content, reasoning, brainstorming, writing, and planning, though the improvements are not quantified.

​

Let's elaborate a bit

The exact version that Bard will be benefitting from is Gemini Pro, the more lightweight version compared to its flagship brother Gemini Ultra, but still more advanced than Gemini Nano, a weaker version of the model meant for Pixel devices, all three of which launched back in December of 2023. The other part of the big update is Imagen 2, also released in December and only available in English, which will give Bard image generation support. An interesting quirk to images created by Bard is the watermark, dubbed ​SynthID​, developed by Google's R&D department DeepMind.

Studio's newest online school platform with ​AI-based​ custom curriculums for musicians hits the shelves​

The rival to MasterClass, Studio, launched its new AI-powered online school on Wednesday for musicians, songwriters, and producers to learn from the very best of the industry. With two pricing options, monthly and yearly, Studio's Music School is for avid musicians to create new songs, get feedback from peers, and access Studio's AI coach with thousands of exclusive lessons taught by roughly 110 top artists. The AI-powered coach featured on Studio's Music School uses GPT-4 to deliver curated lesson plans every month based on the person's interests and available time, with the ultimate goal of always finishing one release-ready song by the end of the month.

​

​Mastercard to introduce generative AI to its systems to increase fraud detection​

Mastercard is next in line to jump aboard the generative AI train, as the company recently announced its plans to tackle fraud by building its own proprietary generative artificial intelligence model. This new advanced model, dubbed Decision Intelligence Pro, will allow banks to determine in real-time whether suspicious transactions are legitimate. Ajay Bhalla, Mastercard’s president of the cyber and intelligence business unit, says that the "majority" of the model is developed in-house and trained on roughly 125 billion annual transactions. On average, the company claims that Decision Intelligence Pro increases fraud detection by 20%.

​

​Apple is inducing excitement in the crowd with talks of generative AI features coming "later this year"​

Tim Cook, the CEO of Apple, confirmed that generative AI features are on the way and will arrive "later this year" during an earnings meeting. Cook teased a wide array of generative AI features, though regardless of how much analysts at the table tried to press him for more details, he never revealed any specifics, claiming that "Our M.O., if you will, has always been to do work and then talk about work, and not to get out in front of ourselves." Despite his reluctance to reveal more, many see this as lining up with Bloomberg's report that claims iOS 18 could be the "biggest" update in the operating system's history, coming most likely this autumn.

​

​Generative AI is getting tested in a limited access run for Google Maps​

​In a Thursday blog post, Google announced that the company would begin a trial run of new generative AI features for its Maps app. The early access release for these features will be available starting this week in the United States for select Local Guides, Google's community of active members who contribute detailed information about different locations. Google claims that the feature should feel "conversational" rather than a regular search experience, adding that the new feature will be able to respond to even the most niche of queries.

​

​Twin Labs develops an AI tool to take over handling repetitive tasks​

A Paris-based startup, Twin Labs, brings an intriguing idea to the table, tackling the issue of dull, repetitive, and yet mandatory tasks by leveraging GPT-4V - OpenAI’s vision-capable GPT model. Twin Labs co-founder and CEO Hugo Mercier said the company opted for a vision-based model because LLMs (large language models) prove themselves to be constantly unreliable by making “the wrong decisions.” Meanwhile, models like GPT-4V understand the feature behind different interfaces, like a button saying “Subscribe,” which gives vision-based models the potential to be efficiently used for long, repetitive lists of tasks. However, the project isn’t there yet and is a work in progress as the two co-founders continue to develop a prototype of this product, backed by a $3m pre-seed fund.

​

​Universal Music Group threatens to pull major music artists off TikTok amid contract dispute​

With its current arrangement with TikTok set to expire on the 31st of January, Universal Music Group (UMG) said in a press release that it would not be renewing its contract with the social media platform, as well as stating plans to cease licensing content to TikTok Music, too. This decision stems from UMG's allegation against TikTok, saying the Chinese media giant is trying to build a "music-based business without paying fair value for music" and a lack of protections against AI-generated music. A TikTok spokesperson fought against these allegations, saying that UMG has "put their greed above the interests of their artists and songwriters." Nevertheless, UMG's decision will cost TikTok licenses to notable artists, such as Taylor Swift, Drake, Ariana Grande, and Billie Eilish.

​

​Yelp to launch a variety of new AI-powered features to its platform​

As part of its winter release, Yelp is launching a small bundle of new features, which include AI-powered summaries of business, number masking for privacy, and a revamped visual feed. The winter update emphasizes the iOS version of the app first, with an Android to follow in the coming months, to show more visual content, including collections from ​Yelp Elites​ and videos posted by businesses, additionally adding a new search experience that will give suggestions before you even begin typing your query.

Artisan - Artisan lets you create AI digital workers that can seamlessly integrate into human teams. With an onboarding process of only 10 minutes, Artisan's digital workers are designed to help your job workflow.

​Dashtoon - Dashtoon is an AI-powered platform designed to simplify and enhance the process of creating comics, regardless of whether you're a veteran or someone new with some curiosity.

​Letterly - Letterly's AI model turns your speech into a comprehensive, well-crafted text, be it a letter, email, or general notes - except it's not a mere transcription. Capture your voice, and let AI do the writing.

​Shakker - Shakker AI is a powerful image-to-image generation model, allowing you to create images, switch up styles, throw together components, or inpaint any part.

A Race Against Time (8-min read)

Dive into the pivotal role of antitrust regulations in shaping the tech-infused future as both U.S. and European enforcers discuss regulations for AI in a race against time.

​

The Singularity Draws Near (70-min watch)

Ray Kurzweil, an American inventor, futurist, and pioneer in artificial intelligence, answers questions from the audience about AI, and the impact and effects this future holds for society.