The Generative Edge Week 17
We check out AI assisted search engines so you don't have to, multi-lingual voice cloning is making strides and video games are about to change.
Welcome to the Generative Edge of week 17. Here’s the gist in 5 bullets:
AI-powered search engines like Phind, Bing Chat, and Perplexity aim to improve upon issues faced by LLMs like hallucination and outdated data.
Bark, a new model by Suno AI, offers generative voice generation with features such as handling laughter, sighing, clapping, multi-lingual support, and active voice cloning.
Generative AI is transforming video game development by making asset creation more efficient, with big players incorporating AI for code, story development, NPC dialogue, and more.
UnrealEngine 5 and Stable Diffusion enable quick and efficient generation of game worlds and art pipelines.
Google is rumored to be working on "Magi," an AI-assisted search engine project.
No major announcements this week, but the generative AI space marches on regardless, and we have some interesting tidbits for you. Let’s dive in!
AI assisted search
Today, we are going to take a quick look at various AI powered search engines. Using a LLM in isolation is very helpful and you can get a lot of value out of that, but you also quickly run into issues (hallucination, data not up to date etc.).
AI enabled search engines promise to solve some of these issues, while providing the same powerful dialog based interface.
Here are some candidates:
Honorable mentions: ChatGPT Search, DuckAssist
Task 1 - Code generation
generate me some python code that will interface with gpt-4
Task 2 - Cooking recipe
give me a vegan recipe that is fast and yields a lot of nutritious food for 2 people
Based on our testing, we think that Phind performs best, followed by Bing Chat and depending on how current or specialized the request, vanilla ChatGPT. Perplexity kind of disappoints, ChatGPT Search is still in closed beta, DuckAssist isn’t really an assistant and Google is entirely absent with their offering.
Bark is a new model released by Suno AI. We’ve mentioned the area of generative voice generation a few times already, but this one has a few new tricks under its sleeve:
This is an open model that you can run yourself!
It also knows how to deal with laughter, sighing, clapping and other more natural components of human vocal communication
It supports multi-lingual voice generation, including code switching (switching to another language in the middle, including adding an accent).
It supports active voice cloning (record a sample with the right inflection, have another voice say it exactly like that)
Check out some of the examples that which are entirely generated:
Hey Martin, es ist wirklich beeindruckend, dass du nur mit deiner linken hand programmieren kannst. It's super cool, but maybe you could type faster if you used your right hand too.
[Sad] My friend’s bakery burned down down last night. [sighs] Now his business is toast.
I got a face for radio and like, what do they say? A voice for — um — for print? [laughs]
Video game worlds
Video game development is an area that will benefit greatly from generative AI. All video games have something in common, no matter if a small indie or a large AAA title that costs 50-100 million Euros to make: they all take a lot of effort and money to make. Asset creation in particular is really expensive, and gen AI is poised to make that process more efficient.
UnrealEngine 5 is one of the most popular game engines out there
Epic (the company making UE5) started pushing into traditional media creation as well and provides the technology that powers virtual sets (sometimes called The Volume)
In the example above, a Stable Diffusion generated texture is applied to a virtual terrain, which means developers can create a (2.5D) game worlds quite quickly with little effort.
Generative AI is being incorporated by all the big players in this space, and it’s useful almost everywhere, from code and story development, NPC dialogue, asset generation, QA, concept testing and more - it’s a big deal.
and what else?
Google is rumoured to work on “Magi”, their own take on AI assisted search.
And that’s it for this week!
Find all of our updates on our Substack at thegenerativeedge.substack.com, get in touch via our office hours if you want to talk professionally about Generative AI and visit our website at contiamo.com.
Have a wonderful week everyone!
Generative AI engineer at Contiamo
Thanks for reading The Generative Edge by Contiamo! Subscribe for free to receive new updates from the world of generative AI.