UPP News: What's new in the AI field? (Part 9)

Subscribe to our newsletter and stay up-to-date with the latest AI news updates.

Source: OpenAI

May 13th, 2024

OpenAI has introduced GPT-4o, a new generative AI model that integrates text, speech, and video capabilities. This model, described as having "GPT-4-level" intelligence, is designed to handle multiple modalities and improve interaction between humans and machines. GPT-4o enhances the ChatGPT experience by enabling real-time voice interactions, nuanced voice responses, and advanced vision capabilities, such as analyzing images and answering related questions. The model supports around 50 languages and offers faster, more cost-effective performance than its predecessor, GPT-4 Turbo.

GPT-4o's rollout will be gradual across OpenAI's products, initially accessible in the free tier of ChatGPT and to subscribers of ChatGPT Plus and Team plans. Voice features will be selectively available due to potential misuse risks. A refreshed ChatGPT UI and a new macOS desktop app are also being introduced, with a Windows version coming later. Additionally, the GPT Store and certain previously paywalled ChatGPT features are now available to free-tier users.

Source: The Verge

May 14th, 2024

TikTok is testing a new feature called "search highlights" that uses generative AI to enhance search results. AI-generated snippets from ChatGPT appear at the top of some search queries, offering detailed information when clicked. This feature is limited to certain queries for now. Additionally, TikTok is experimenting with "Search Highlights" that are not labeled as AI-generated. These updates aim to improve in-app search capabilities, leveraging the trend of younger users using TikTok as a search engine for recommendations and information.

Source: TechCrunch

May 9th, 2024

Call centres are rapidly adopting automation, with the global contact centre AI market expected to grow from $2.4 billion in 2022 to nearly $3 billion by 2028. About half of the contact centres plan to implement AI within the next year to reduce costs and improve efficiency. Retell AI, led by Evie Wang, provides a platform for AI-powered voice agents that handle tasks like appointment scheduling using large language models and speech synthesis. While Retell's synthetic voices lack some realism, they are reliable for scripted interactions. Despite challenges with complex queries and potential inaccuracies, companies like Retell remain optimistic about the future of conversational AI in call centres.

Source: TechCrunch

May 8th, 2024

Google DeepMind has unveiled a new version of AlphaFold, its groundbreaking machine-learning model for predicting protein structure and behaviour. AlphaFold 3 boasts improved accuracy and the ability to predict interactions with other biomolecules, enhancing its versatility as a research tool. This release marks a significant milestone in computational biology, as accurate protein structure prediction is crucial for understanding biological processes. Despite ongoing technical challenges, AlphaFold and similar models have revolutionized protein structure prediction, accelerating research across various scientific fields.

AlphaFold 3's advancements enable the simulation of multiple molecules simultaneously, offering researchers unprecedented insights into molecular interactions. Google DeepMind's decision to offer AlphaFold Server, a free web application for non-commercial use, democratizes access to this powerful tool.

However, concerns remain about the proprietary nature of AI models and the lack of transparency in their training processes. While AlphaFold's widespread availability benefits the scientific community, some advocate for greater openness in sharing research findings and data. Nonetheless, AlphaFold's impact on drug design and molecular biology research is undeniable, promising significant advancements in rational drug design and accelerating scientific discovery.

Source: TechCrunch

May 7th, 2024

Meta is launching an expanded set of generative AI tools for advertisers, building upon its initial AI features introduced last October. Advertisers can now request full image variations, allowing for AI-generated ideas that update the photo's subject or product being advertised, in addition to creating different backgrounds. While this feature enhances creative possibilities, it also raises concerns about the potential for advertisers to deceive consumers by showcasing non-existent product variations.

To address these concerns, Meta has implemented guardrails to filter inappropriate or low-quality content generated by its AI models. Despite the precautions, there's still the possibility of misuse, prompting Meta to stress-test the feature with internal and external experts. Advertisers will soon be able to provide text prompts to tailor image variations, further enhancing customization options.

In addition to image expansion, Meta is introducing text overlay features and allowing advertisers to add text overlays with popular font types. These updates aim to streamline the creative process across different surfaces, such as Reels and Feed, reducing the need for manual adjustments. Meta's AI capabilities extend beyond images to generate alternate versions of ad headlines and primary text, with plans to incorporate brand voice and tone in future updates. These generative AI features will be globally available to advertisers by the end of the year.

Source: apple.com

May 7th, 2024

At Apple's recent iPad event, the tech giant highlighted AI technologies integrated across its device lineup, teasing advancements to be fully unveiled at its Worldwide Developer Conference in June. The company emphasized the AI-powered features, powered by the M4 chip, which boasts a dedicated neural engine for accelerating AI workloads. Examples included visual lookup, subject lift, and live text capture, showcasing Apple's commitment to integrating AI into its devices. Apple showcased its hardware advancements, such as the new M4 chip, delivering up to 50% faster performance and featuring next-generation ML accelerators.

While Apple didn't specify AI use cases during the event, it hinted at enhanced AI capabilities soon to be available to developers through iPadOS. The company highlighted existing AI features in apps like Logic Pro, introducing AI-powered session players and machine learning-backed plugins. Although Apple's AI announcements were not groundbreaking, they underscored the company's focus on leveraging AI to enhance the consumer experience. While hardware updates took centre stage at the event, expectations are high for Apple's AI plans at WWDC, where rumours suggest announcements regarding AI-powered Siri or potential partnerships with AI providers like Google or OpenAI.

Source: TechCrunch

May 12th, 2024

Anthropic, an AI startup, is now permitting minors to use its AI systems through third-party apps, provided developers include safety features and comply with regulations like COPPA. This shift reflects a broader trend in the AI industry towards child-focused applications, while also addressing concerns about safety and ethics. However, challenges remain, prompting calls for clearer guidelines and regulations from organizations like UNESCO. Anthropic's approach combines technological measures, regulatory compliance, and educational initiatives to promote responsible AI use among young users.