We have to end it with a bang‼️

For those who celebrate, I hope your Christmas was great and the spirit is still alive and well. This is my last AI In The Middle until 2023! Calin Drimbau of broadn will be publishing a special edition of the newsletter on Thursday that you won’t want to miss.

What do we have for you today?

It’s provocative: AI-generated pornographic media is causing a stir this week.It’s sensual: my thoughts on a vision for generative AI breakthroughs and its multisensory potential.

NSFW AI Generated Media 🍑🍆

AI-generated porn🌽 isn’t new nor a secret. However one startup, Unstable Diffusion, is shaking the table this week.

The Drama

Unstable Diffusion was able to successfully raise $56,000 via Kickstarter (a little over 2x its goal of $25k), only to have it all stripped away on December 21st by Kickstarter due to its change of view on AI content. Other crowdfunding/membership platforms have also placed a block on Unstable Diffusion like Patreon.

The sudden change of heart by Kickstarter was surprising to many. For others not so much. The company cited concerns over whether AI generated content is created without artists’ consent, exploits specific communities, or places anyone at risk of harm. These views are common and have been brought to the forefront for many months now.

The Argument Over Consent and Ethics

There’s a rise in Anti-AI sentiment especially among artists.

For some, they have issues with these newly public AI systems imitating their likeness and styles without consent, proper credit, or compensation. Some feel ecosystems like Unstable Diffusion are cesspools for collecting controversial media in a central place. They attract malicious users to commune and develop their harmful ‘skills’ to paraphrase Abhishek Gupta, Founder and Researcher at the Montreal AI Ethics Institute, told TechCrunch.

Of course there are the repeated callouts as well. These AI models aren’t diverse, they amplify biases and defaults already seen in the adult entertainment industry, and they can exacerbate traumas for content moderation teams who have to review and remove offensive content.

Popular adult content platforms like OnlyFans have taken stances on Gen AI content too.

For now, OnlyFans allows content of all kinds, including AI-generated media, with the understanding that the person involved in the content is verified, consenting, and of-age. There’s obviously a strict no-no on deepfakes though. To Unstable Diffusion’s defense, the startup shared it “strictly only allows fictional and law-abiding generations, for both SFW and NSFW content, on our Discord server.”

The Resilience

The Kickstarter rug-pull isn’t keeping them down though. Unstable Diffusion CEO, Arman Chaudhry, plans to pursue venture funding in 2023.

In the meantime, Unstable Diffusion has already enabled supporters to donate directly to the startup via their website. As of writing, Unstable Diffusion has raised over $25k, beating their initial fundraising goal.

Founded only in August 2022, Unstable Diffusion is actually a subsidiary of Equilibrium AI (a company also founded by Arman Chaudhry) and therefore has its eyes set on avenues outside of in-house projects.

Their plans include creating and fine-tuning their own models for special use cases as new brand and product spin offs. One example being Waifu Diffusion, Unstable Diffusion’s model fine-tuned on anime images for AI anime porn🌽 (hentai). Another contender in this specific space of adult anime content being YoDayo also seeking venture funding in Q1 2023.

The Road Ahead

What’s for certain, Unstable Diffusion and other AI Adult Content Generators have a steep uphill battle. Moderation, safety, equity, and privacy are all top of mind. The last thing any of these types of startups want is a situation like MindGeek (PornHub’s parent company) in 2020. MindGeek saw major payment processors revoke their support and accessibility for the adult platform after PornHub was revealed to have child pornography, revenge porn🌽, and sex-trafficking videos hosted on its site.

What Can Come Next for Generative AI?

My thoughts as 2023 is around the corner, what could we imagine comes next with Generative AI?

With 2023 approaching, I believe the unicorn AI startups positioned to win in the next decade will be multisensory. The prospect of Gen AI becoming multisensory (”multimodal”) is on my mind and fascinates me. We have now seen systems like DALL-E, ChatGPT, Midjourney, etc., take the tech world by storm in a matter of weeks. Because when their results are good, they're really good. And they only leverage one sense.

Multisensory learning is a theory that we learn better when taught using multiple senses. It considers VAKT (visual, auditory, kinesthetic (movement), and tactile (touch)) senses—in some instances, taste, olfactory (smell), and balance are added too. Translating this concept to the text-to-image/video/text outputs live today imagines generative AI systems creating new substantial outputs and becoming “even smarter.”

Imagine a Gen AI system which incorporates haptics (kinesthetic/touch) as input. This could be a game changer as to how machines interact and understand our physical world. It can affect the ways we work and what data we’re able to retrieve as a result. Ideas that come to mind:

• In HealthTech, multimodal AI could potentially craft highly personalized, holistic recommendations and diagnoses by cross-pollinating haptic data from biosensors, patient scans and imaging, test recordings, and medical histories.

• In Luxury and Beauty, multisensory AI could generate uniquely curated fragrances, attire, marketing, or even skincare by coupling various data inputs like visual, olfactory, textual, and tactile, collected by comprehensive datasets. Seasonal drops, "personal lines," sustainable research, fashion weeks, and fast-fashion can be the first to benefit from these advancements.

• In Mixed Reality (MR), ideas of multimodality enhancing AR/VR experiences and technologies with, and through, multisensory activation as Gen AI becomes more nuanced and sophisticated can be mind-blowing for gaming, media, and entertainment.

• In Synthetic Data, multisensory AI can make for richer, more dynamic inputs and outputs. This could unlock new ways machines understand our world (producing more complex or superhuman outputs) or even how we understand our world.

I’m no psychic so it’s difficult to say with conviction what’s possible and what’s not. However, we can all agree generative AI is disruptive and here to stay.

I’m patiently waiting the day we get an immersive Pokémon game in VR so I can live out my childhood dreams. I’m only sort of joking.

Gen AI Deals that make your eyes (and mouth) water 💰

👉 Deepfake Studio, Deep Voodoo, reels in $20M before the new year

What caught our eye? 👀

From the creators of South Park and Kendrick Lamar’s sensational music video for “The Heart Part 5”, Deep Voodoo specializes in deepfake curation among other generative AI capabilities. With the $20M raise, led by Connect Ventures, Deep Voodoo can scale operations and offer its services to other businesses.

New developments to spam your #random Slack channel 💬

• The 2022 State of AI Compute Report Index

• ICYMI: Quora launched Poe, an AI chatbot like ChatGPT, and joins the wave like Notion AI

• Think your art’s been used to train AI and you want it removed? Here’s Have I Been Trained, a tool developed by Spawning granting artists visibility into knowing if their work appears in popular datasets and how to opt out.

Things to learn when you need a raise

• Pitchbook: VCs try to parse through the 'noise' of generative AI

• Navigating Opportunities and Challenges of ChatGPT and other LLMs

Want access to the Generative AI investors database? Share AI In the Middle (below) to receive it! ⬇️

Before you go

Beloved fandoms meet Gen AI: Star Wars, but make it 80’s Japanese anime. Lord of the Rings, but make it Studio Ghibli.

That's it for today folks! Happy Kwanzaa & New Year, my next letter will see you in 2023 🌟

— Matthew J. Sánchez (@matthewjsanchez)