The Generative AI hype cycle is at its peak. OpenAI, creators of GPT-3 and Dall-e 2 among other AI innovations, is about to make their technology more widely available by opening an API (in beta) for businesses and developers to incorporate generative AI within their products. In the announcement, OpenAI revealed last week that more than three million users now produce four million images every day using DALL-E. Readers of VC Cafe haven seen my various posts on the potential of creative automation, the opportunities it brings, and the impact AI will have on jobs.
Currently, we’re in the ‘playful’ stage of generative AI. People are trying things out and enjoying the ‘magic’ feeling of creating images, text and video from text prompts. The use cases in practice though, are mostly limited to replacing stock photos with AI generated images. I often create images to go along with my blog posts using Dall-E as an example. Microsoft went a step ahead and integrated Dall-e in the Office Suite, enabling people to easily integrate AI generated images in their presentations.
We are starting to see more specific models being trained for better use-case-specific results. But with Stable Diffusion, which offers an open source generative AI API already, we’re starting to see developers training the generative AI algorithms to perform specific tasks. For example, in the case of Israeli startup Strmr, people pay $3 to train the system with their profile pictures to generate cool portrait pictures with AI.
The next step, I believe will be generative AI to perform specific tasks in the world of work. We already have a glimpse of it in text: Jasper.ai (which recently secured $125M in funding at a $1.5 billion valuation) and Copy.ai, are already doing a pretty good job generating text for marketing/ ad copy.
Startups are pouncing on the opportunity. As APIs become more widely available, founders can focus on the application layer generative AI, without the burden of developing their own model. Can startups build a venture scale model using open source tech or someone else’s API? there are plenty of examples that prove the answer is yes.
But what’s coming next? Could AI help us better do our jobs, no matter what you do? I’m not talking about completely replacing humans, but actually enhancing performance, assisting in quality control and enriching our frame of reference. I believe that AI use cases will become more vertical and train to specific tasks, deigned to help each and every role in a company….
Forrester predicts that 10% of Fortune 500 companies will use AI in 2023 to generate content. Rowan Curran, the AI/ ML Forrester analyst said:
The pace of AI change is happening so fast, with such a broad adoption of large language models across different use cases, that this prediction already feels almost out of date.
I should probably have scaled it upwards. I think maybe I would have revised it to 10% of Fortune 500 workers will use these tools, because to me that speaks to the way this AI trend is evolving — I think it’s going to bubble up from below as much as it’s going to come down from above.”
It’s not hard to imagine what this could look like by using a simple idea generation exercise:
- Developer + AI – coding autocomplete tools like Github’s copilot or Israel’s Tabnine, AI for QA
- Designer + AI – Text to 3D, text to image for designs, automated brand assets creation, etc
- Product Manager + AI – GPT-3 for product specs, AI recommendations for features, automated A/B testing
- Marketer + AI – GPT-3 for blog posts and ads, text to image for product visualisations
- Business Development + AI – text to video, speech to speech to localise messages in the local language, AI for knowledge base (like pragma.ai)
- HR + AI – text to video for onboarding and recruiting, GPT-3 for recruiters personalised messages
- CFO + AI – AI for book keeping, predictive analytics for revenue
And of course, non-tech jobs, from oil and gas to automotive, manufacturing, education, biotech, health, etc. The potential is immense.