A Chat with Steve Harris, CEO at Synthetic Data Platform Provider: Mindtech

We are a UK based technology company, whose vision is to improve the quality of AI vision systems. We do this by enabling our customers to create relevant, privacy-compliant and bias-free datasets – traditionally hard to compile – that can be used to better train AI.

Our customers create the data they need using our Chameleon Platform. With Chameleon, they can create photorealistic virtual worlds and then simulate real-world events to create the thousands of high-quality, annotated images they need to teach their AI network how to respond to different situations.
Mindtech, Synthetic Data Platform Provider, Announces Strategic Partnership  With Appen and Raises $3.7 Million | Business Wire

How did you come up with the idea for the company?

The idea for Mindtech stems from our shared belief that AI development should be inclusive, and open to all.

Our world-leading Chameleon Platform is built to address the growing demand for better diversity in datasets. Just this month we launched an update which allows users to rapidly create millions of variations of the ‘synthetic actors’ they include, meaning age, race, fashion and regional characteristics can be changed automatically. This is essential for creating AI systems that better understand how humans interact with each other and the world around them, without being influenced by the customer’s own innate bias.

Chameleon also fully supports the newly introduced Monk Skin Tone Scale to help diversity in AI data training. AI systems historically struggle to recognise darker skin tones, but The synthetic actors included on Chameleon can be configured to represent a broad sweep of skin tones to help bridge the gap.

These improvements make life easier for data scientists training AI systems, too, who can spend 80% of their time on data management – cleaning, labelling and annotating for the data itself to be usable. The increased use of synthetic data provides them with diverse, high-quality, annotated images that will produce better results and cut management time.


How has the company evolved over the last couple of years?

The company has grown significantly over the past years. Last year, we updated Chameleon to enhance the user experience, simplifying the creation of complex scenarios, with comprehensive new data discovery, bias and diversity management tools. New automated real-world behavioural models built into the simulator ensure the required training datasets can be built with ease.

We’ve also recently announced a strategic investment round led by Appen, the global leader in data for the AI Lifecycle. This new investment round, including participation from Appen and existing Mindtech investors, follows a $3.25 million funding round in July 2021 and will be used to support the company’s rapid growth.

In addition to the investment, Appen and Mindtech have formed a commercial partnership to provide a range of real and synthetic images and associated data and metadata annotation services to the market.

Appen’s leadership in labelling real-world data combined with Mindtech’s end-to-end platform for the creation and management of synthetic data will accelerate the development of more accurate AI systems.

What can we hope to see from Mindtech in the future?

Our work so far has eased the bottleneck in AI training, fueling the hyper-automation of AI itself by allowing companies to generate synthetic training data to make their AI systems smarter at a rate that’s 50x faster than using real-world data alone.

We’ll continually improve Chameleon too, as we have done with our recent synthetic actor update to help make AI systems more diverse and less biased, and our partnership with Appen to be able to top up a high volume of synthetic training data with some real-world data – generally considered to be the most effective approach.

Of course, the metaverse is the opportunity on every technology company’s mind at the moment, and Mindtech will have a role to play. We are still figuring out exactly how AI vision systems might work in the metaverse, and what the requirements for synthetic data to train them will be. But, there is the potential for synthetic data to be a prominent tool for metaverse companies when you consider that machine learning models for our real-world are being trained on computer-generated data to an extremely high standard.

That same synthetic data is going to be even closer to a digital metaverse, so there’s no reason why it can’t be used and bring about even better performance than it does now.