AI Weekly Insights #11

AI in 2024: Predictions, Copilot, And Multimodal AI

Happy Sunday!

Happy New Year from the 11th Edition of “AI Weekly Insights”! I took a break for the holidays, but am back to deliver the weekly insights around AI! Because I took some time off, this edition will be filled with things that were missed and may not only be for the last week. We will continue our regularly scheduled program next week!

(P.S. Click the story’s title for more information 😊):

  • What’s New: TechCrunch has made predictions for where AI will go in 2024. Some of the highlights are below:

  • Top 3 Predictions:

    • Agents + Generative Audio & Video: AI Agents allow AI to act autonomously to complete a given task. While image creation was big in 2023, I expect video and music to also see some further advancements in 2024.

    • Apple Enters The Race: Apple has long used AI in their products, but will often refer to it as Machine Learning. I expect tools like Siri will start to use an LLM to power its responses.

    • The 2024 Election will be Worse with AI: Bot accounts and fake blogs will be used to constantly generate nonsense and confuse people. This is already a problem with text and can be much worse with images and videos.

  • Why It Matters: I fully expect multimodality to become a big thing in 2024, where you’ll have AI systems that can easily understand text, images, audio, video, etc. The implications of this technology being used in the upcoming election are worrying. I could see companies or technology being created to mitigate this in the coming months.

Image Credits: Getty Images

  • What's New: OpenAI will finally release the store for users to sell their custom GPTs after delaying it last year.

  • The GPT Store: GPTs (announced in November 2023) allowed users to create a custom version of ChatGPT, tailored for a specific use case. Users were able to share these GPTs with others by sharing a link. This will allow GPTs to be offered more widely to customers and could open up business opportunities for individuals.

  • Why It Matters: GPTs are extremely useful for having a more specialized version of the chatbot. While you could find ones shared by others if you knew where to look, having a centralized store will make it much easier for people to try these out. Exceptionally useful ones could become staples for anyone looking to use the service.

Image Credits: Justin Sullivan / Getty Images

  • What's New: Over the holidays, Microsoft quietly launched its Copilot app on iOS, iPadOS, and Android.

  • Mobile Copilot: The app gives users access to Copilot which operates similarly to ChatGPT. Users can use the assistant for email drafts, composing stories, summarizing text, and much more. Using the capabilities powered by DALL·E 3, users can also generate a wide variety of images in a variety of ideas and styles.

  • Why It Matters: Enabling users to access Copilot on mobile greatly increases its use and access to people worldwide. Using Copilot also allows users to access the technology using GPT-4 for free (which OpenAI charges for with ChatGPT).

Image Credits: Microsoft

  • What's New: Microsoft is adding a new physical key to keyboards that would invoke Copilot in Windows.

  • Copilot: Yusef Medi, executive vice president at Microsoft, writes that this “marks the first significant change to the Windows PC keyboard in nearly three decades.” The first PCs that will ship with this key will likely come in February after this year’s CES (Consumer Electronics Show) in Las Vegas.

  • Why It Matters: I’ve covered that Microsoft may be looking at 2024 as the “year of AI PCs” and a bit about how Copilot is poised to change the way we use our computers. This is further evidence of the company’s push to empower people to participate in the use and development of AI.

  • What's New: Midjourney has released version 6 of their image creation model, currently available in an alpha testing state.

  • New Prompts: Prompting with the new version is significantly different than the current version, and there will be a slight “relearning” of prompting. The new model is much more sensitive and prompts like “photorealistic, 4k” are considered junk.

  • Why It Matters: The new model is MUCH better at generating different styled images while getting better at generating two of its biggest issues: hands and text. You can see the difference below with my prompt, “photograph of an African-American woman facing the camera”.

Generated with Midjourney v5.2 (Current Version)

Generated with Midjourney v6 (Alpha version)

  • What's New: Spotify is testing a feature that would allow users to create new playlists based on text prompts and AI.

  • AI Playlists: The launch of AI-driven playlist creation would fall in line with Spotify’s use of the tech with last year’s launch of an AI DJ that offered more personalized playlists with commentary. The CEO notes that AI could further be used to summarize podcasts or even automatically create audio ads in the future.

  • Why It Matters: The company tests a variety of features all of the time but has shown to have taken a big interest in AI over the past few years. As someone who creates (and listens to) music, the use of this type of technology in this space is always cool.

Image Credits: Jakub Porzycki / Getty Images

Thank you to all who subscribed and helped make 2023 fun for finding and sharing things in AI! Without you, I would not be able to do this ❤️. Here is to 2024 being another amazing year, I can not wait to see what is in store.

Share these insights with pals who'd love a weekly AI digest 🔄. Drop a comment or hit reply to share your thoughts. See you next Sunday!

Warm regards,

- Kharee