MED-AI: Omnicom, Google, Meta
In AI news this week, the CEO of OpenAI Sam Altman found himself in hot water. As mentioned in a previous MED-AI roundup, the chief executives of the company, along with hundreds of other AI companies and technology experts previously signed a one-sentence statement saying that “mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.” Altman has repeatedly urged governments to move towards global AI regulation.
But according to documents obtained by TIME this week, OpenAI has lobbied for significant elements of the EU’s AI Act to be watered down such that it would reduce its own regulatory burden. The publication reported that, in several cases, OpenAI proposed amendments that were later made to the final text of the EU law. The AI Act, which was approved by the European Parliament earlier this month, represents the most comprehensive AI legislation in the world.
In more optimistic news, Meta scientist and one of three “godfathers of AI” Yann LeCun believes that some experts’ fears of AI posing a threat to humanity are “preposterously ridiculous”. He continued: “Will AI take over the world? No, this is a projection of human nature on machines.” LeCun also expressed his doubt that AI would pose a permanent threat to jobs.
Hear that? You’re going to have to continue doing your jobs – at least for a while – and MED-AI is here to give you your weekly round up of everything AI related that you need to know.
Omnicom
Omnicom announced that it is integrating Google’s generative AI models into its ad tech platform. Through the integration, brands will be able to access generative text and image capabilities, “expediting the content development process, providing enhanced efficiency and creativity” according to the companies.
Omnicom will leverage the capabilities of Google Cloud’s Vertex AI, specifically its foundation models such as PaLM 2 and Imagen. PaLM 2 is Google’s large language model for text and Imagen is similar to DALL-E, for generating images. Imagen enables organisations to generate and customise ‘studio-grade’ images at scale from input text.
The partnership represents the first commercial application of Google’s generative AI technology.
In more Google news this week, the Alphabet company announced new generative AI features added to its online shopping experience. The “try on” will allows users to understand how apparel will fit on them regardless of their body size, overlaying apparel on top of images of diverse models. The Alphabet-owned company said the service is powered by a new image-based AI model that it developed internally, and will soon release a new research paper detailing its work.
Google also announced news way to research travel destinations and map routes on the search engine using generative AI.
Meta
Meta revealed it has developed a new tool called “Voicebox” which it says can generate convincing speech in a variety of styles. According to the company, the solution can create new voices from scratch or use a sample, and works across six languages. Unlike other speech systems that require training for each task, the tech giant says that Voicebox can just be given raw audio and a transcription. The company said it will not release the program for fear of abuse.