AI Industry Updates: February 2024

Published 14 February 2024

Author
Emilia Morano-Williams
4 min read

This February, the artificial intelligence (AI) sector is pushing ahead with regulations, certifications and best practices intended to limit the influence of deepfakes and instil ethical behaviours. Meanwhile, consumers will appreciate the influx of lifestyle-focused tools, designed to streamline their digital lives.

Bolstering AI Guard Rails

Bolstering AI Guard Rails

  • Ethical AI Certification: American non-profit Fairly Trained introduced its Licensed Model certification scheme, which identifies companies that use legally obtained data to train their AI algorithms. To become certified, businesses must use data explicitly licensed for training purposes, in the public domain, licensed appropriately or belonging to them. The certification fee ranges from $500 to $6,000 (depending on company size) and allows businesses to display the Fairly Trained watermark.

    Currently, many large language models (LLMs) – including California-based OpenAI’s ChatGPT – use non-licensed content for training. In December 2023, The New York Times sued OpenAI for taking its articles without permission to train its chatbots. While OpenAI asserts that the cross-posting of digital content means it’s impossible to eliminate copyrighted material from training data, discussions around AI’s copyright implications are likely to accelerate in 2024, making certification schemes like Fairly Trained’s increasingly relevant.

  • New Tools Detect Artificially Generated Content: Between deepfake porn of popstar Taylor Swift and a deepfake phone call of US president Joe Biden before the New Hampshire primary election, there’s a growing urgency to limit unsavoury use of AI. The US government prohibited AI-generated robocalls in response, but more steps are needed.

    Meta advocates for metadata to identify text, images or videos produced with AI. According to Meta, this would allow media companies to easily determine when AI-generated content is shared and automatically label it accordingly. OpenAI is incorporating metadata, alongside a visible watermark, on all images generated by its Dall-E 3 chatbot.

    Targeting text, researchers at the University of Maryland have developed Binoculars, an AI detection tool with a 99.9% accuracy rate. While the tool can identify plagiarism in academic papers, the researchers also noted that it could help recognise AI-created content on social media or news platforms.
  • Ethical AI Certification: American non-profit Fairly Trained introduced its Licensed Model certification scheme, which identifies companies that use legally obtained data to train their AI algorithms. To become certified, businesses must use data explicitly licensed for training purposes, in the public domain, licensed appropriately or belonging to them. The certification fee ranges from $500 to $6,000 (depending on company size) and allows businesses to display the Fairly Trained watermark.

    Currently, many large language models (LLMs) – including California-based OpenAI’s ChatGPT – use non-licensed content for training. In December 2023, The New York Times sued OpenAI for taking its articles without permission to train its chatbots. While OpenAI asserts that the cross-posting of digital content means it’s impossible to eliminate copyrighted material from training data, discussions around AI’s copyright implications are likely to accelerate in 2024, making certification schemes like Fairly Trained’s increasingly relevant.

  • New Tools Detect Artificially Generated Content: Between deepfake porn of popstar Taylor Swift and a deepfake phone call of US president Joe Biden before the New Hampshire primary election, there’s a growing urgency to limit unsavoury use of AI. The US government prohibited AI-generated robocalls in response, but more steps are needed.

    Meta advocates for metadata to identify text, images or videos produced with AI. According to Meta, this would allow media companies to easily determine when AI-generated content is shared and automatically label it accordingly. OpenAI is incorporating metadata, alongside a visible watermark, on all images generated by its Dall-E 3 chatbot.

    Targeting text, researchers at the University of Maryland have developed Binoculars, an AI detection tool with a 99.9% accuracy rate. While the tool can identify plagiarism in academic papers, the researchers also noted that it could help recognise AI-created content on social media or news platforms.

Assistants to Coaches: Expanding AI’s Lifestyle Role

Assistants to Coaches: Expanding AI’s Lifestyle Role

  • Google’s Upgraded AI: Google is expanding the use of its Gemini In name, it will replace the company’s Bard chatbot – the search giant’s answer to ChatGPT. More significant is the introduction of Gemini’s paid tier, which enables subscribers to access the AI directly through apps like Gmail and Google Docs. Gemini is also replacing Google Assistant on Android phones. The company claims this will allow for more conversational help and a step towards a “true AI assistant”.

  • Automating Digital Tasks: Reports suggest OpenAI is working on an AI tool that would automatically execute complex software tasks for a user. Described as an agent, it could independently create travel itineraries or make restaurant reservations, among other use cases.

    Consumers can test digital automations with the Arc Search, from The Browser Company (US). Arc deploys AI to condense web searches into a single step. After a user enters a query, the AI searches for answers, identifies the relevant details and shares them in a new tab alongside a source.

  • Shop-for-Me Bots: As we saw at NRF, Retail’s Big Show 2024, retailer-launched AI chatbots are aiding online product discovery. See Amazon’s Rufus AI shopping assistant, which parses through the website’s stock to identify products that meet specific user needs. The bot is trained on customer reviews and general web data, which enables it to answer general questions like “What is the easiest type of coffee maker to clean?” or “What’s the difference between a drip and pour over coffee maker?”

  • Computerised Coaches: AI chatbots are positioned as a cost-effective alternative to expensive wellness coaching services. The Tomo life-coach chatbot (from the makers of American AI companion Replika) guides users through 250 wellness practices, including yoga, meditation and positive affirmations. It can suggest specific modalities that meet users’ most pressing needs, determined via chat. For more, see 10 Wellness Trends to Watch: 2024.
  • Google’s Upgraded AI: Google is expanding the use of its Gemini In name, it will replace the company’s Bard chatbot – the search giant’s answer to ChatGPT. More significant is the introduction of Gemini’s paid tier, which enables subscribers to access the AI directly through apps like Gmail and Google Docs. Gemini is also replacing Google Assistant on Android phones. The company claims this will allow for more conversational help and a step towards a “true AI assistant”.

  • Automating Digital Tasks: Reports suggest OpenAI is working on an AI tool that would automatically execute complex software tasks for a user. Described as an agent, it could independently create travel itineraries or make restaurant reservations, among other use cases.

    Consumers can test digital automations with the Arc Search, from The Browser Company (US). Arc deploys AI to condense web searches into a single step. After a user enters a query, the AI searches for answers, identifies the relevant details and shares them in a new tab alongside a source.

  • Shop-for-Me Bots: As we saw at NRF, Retail’s Big Show 2024, retailer-launched AI chatbots are aiding online product discovery. See Amazon’s Rufus AI shopping assistant, which parses through the website’s stock to identify products that meet specific user needs. The bot is trained on customer reviews and general web data, which enables it to answer general questions like “What is the easiest type of coffee maker to clean?” or “What’s the difference between a drip and pour over coffee maker?”

  • Computerised Coaches: AI chatbots are positioned as a cost-effective alternative to expensive wellness coaching services. The Tomo life-coach chatbot (from the makers of American AI companion Replika) guides users through 250 wellness practices, including yoga, meditation and positive affirmations. It can suggest specific modalities that meet users’ most pressing needs, determined via chat. For more, see 10 Wellness Trends to Watch: 2024.