Guest: Kent Keirsey (LinkedIn)
Company: Invoke AI (Twitter)
Show: Let’s Talk About AI

Invoke AI provides an integrated set of generative AI tools to help professional creatives augment their work, streamline their workflows, complete asset creation faster using their own style, and maintain ownership of their models and intellectual property.

In this episode of TFiR: Let’s Talk About AI, Kent Keirsey, Founder and CEO of Invoke AI, shares his insights on the impact of generative AI in the creative space and how Invoke AI is helping individual artists and enterprises co-create with this technology.

Backstory:                                                                         

• Invoke AI started out as an open-source project with a large community of architects, ad agencies, artists, game designers, graphic and product designers.
• It integrates with openly licensed diffusion models, like Stable Diffusion, which allow artists to opt out of having their works included in the training sets.
• It has been one of the fastest-growing open-source software repositories since its inception in 2022.
• Invoke AI built tools that enabled professional creatives to co-create with AI rather than simply throwing a text prompt and hoping to get something useful from the model.
• Earlier this year, they spun up a commercial arm to work with enterprises, especially in media and entertainment, who are looking to accelerate their 2D asset creation pipelines.
• Keirsey started out in technology consulting at Ernst and Young, has worked on B2B/B2C products, two-sided marketplaces, D2C SaaS products, and enterprise SaaS, building good products for customers and identifying how to apply technology to customer problems.

How Invoke AI helps artists and enterprises:

• It’s important to ensure that people have access to the technologies, especially if their field is being disrupted by it.
• The Community Edition of its core generative AI tool can be downloaded and run on your computer for free.
• With more advanced professional workflows, you are no longer just throwing text at a model and hope you get an image. You can actually co-create with it. You can draw an image and pass that in. You can train it on your style, you can use rendering tools, whether you’re an architect doing something in Revit, or a 3D artist working in Blender or Maya.
• You can use multimodal inputs to help guide the AI creative process in realizing your creative vision.
• Invoke AI’s goal is to empower the entire world with good technology. They are helping individual artists who need to use these tools by teaching them how they can train their own model and help them monetize and distribute that model to potential customers. They are helping enterprises understand the value of these models, train their own custom models, and deploy that studio across their teams.

On what’s ahead:

• Large language models (LLMs) have been one of the biggest drivers of generative AI and there are open models that you can download, fine-tune, and use locally either as an end user or as an enterprise.
• When artists saw what the systems with generative AI can do, the initial response was fear of being replaced.
• Now, a lot of artists are starting to see how this can empower them. These models don’t really have a good understanding of their style unless they’re trained, and the only people who can train them well are the artists themselves who have those high-res assets needed to train those things.
• New jobs will be created at enterprises for these types of roles.
• The individual creator now has the capability to do an end-to-end production.
• Music, movies, and video games will be created by small teams that are now capable of doing triple-A quality work with that small team, because AI is helping them do that.
• The creative world will completely change. There will be a creative renaissance with gen AI.

On copyright issues:

• It’s important to separate how these models are trained, how they are distributed, and then how copyrights are obtained on the generated outputs.
• Invoke AI has a very strong perspective that it is not a violation of copyright when you’re training these large models. Stable Diffusion is trained on 5 billion images.
• Training is not copying the image. It is learning those patterns between images so that it can understand what the color red is, what a hat looks like, or what a sunflower might look like. Patterns emerge from looking at multiple images and any individual image is less important than the broad pattern that’s being identified by the machine learning model.
• People are concerned because they don’t want their style easily replicated by somebody who’s going in and typing a prompt. Some issues may arise when distributing models that are intended to recreate the likeness of an individual or recreate their style and commercially compete with them.
• The reality is if they actually try this technology and try to recreate their own style, they won’t be able to without a customized fine-tuned model.
• Invoke AI believes that that type of specialization should take place with the artist, i.e., they should be able to train their own model, they should be monetizing and licensing that model out if someone wants to recreate work in their style.
• It’s nuanced because this is a new age of intellectual property. It’s important for artists to understand the intellectual property that they potentially could own if they approach this correctly.

On Hollywood and Gen AI:

• AI is going to be at the center of how content is created in the future.
• The future of the entertainment ecosystem is going to be creator centric, i.e., it’s going to be those people who can harness the new tools to create those magical experiences.
• Big studios used to be the enabler of creating your vision as a director, as a storyteller. With these new tools, the budget can be a lot lower, creators can tell the story the way they want to tell the story.
• Invoke AI wants to help evolve that ecosystem.
• Creators can leverage these new technologies to realize their vision, but they need to understand the intellectual property that they own and not give up that intellectual property to anyone else. They need to understand how these models are trained. They need to understand how they can optimize those models for their own creative purposes. And they need to be able to deploy that and use that in their creative process.

On “open source” vs. “openly licensed”:

• Open source is probably the wrong label when talking about models, because the models are not really source code.
• The model has a license that you must follow to use it and most of the licenses don’t necessarily map to the licenses we’re familiar with in the open-source ecosystem.
• There’s the code to run these tools and run inference on the model that should be open source. But the models themselves, from Invoke AI’s perspective, should be openly licensed.
• Openly licensed means, barring restriction for abuse or misuse, models should be free to use for anyone and free to create derivatives of those models. It’s important that you can create derivatives of those models because it allows people to fine-tune and customize those. From an enterprise perspective, you want assurance that if you fine-tune or customize that model for your business, it is yours.
• You’re not stuck behind a commercial license agreement. That way, you can actually build this as a core capability of your business rather than something that you’re just renting from a large tech company.

Current trends in AI research:

• There is a strong open-source movement that is moving AI research forward.
• Companies like Hugging Face are advocating for openly licensed and open-source AI research.
• Still, commercial licenses are being applied to some open-source products, meaning, they are no longer really open source.
• There are a lot of openly licensed models being released, even if some of these larger providers are closing some off with commercial licenses.

This summary was written by Camille Gregory.