Startup Of The Week

Startup Of The Week: Mind AI

South Korean startup Mind AI is developing an artificial intelligence engine that converts natural language into a new data structure to perform human-like reasoning, offering what it says are significant advantages over large language models (LLMs) provided by Big Tech companies like OpenAI, Microsoft ,and Google.. The Mind AI engine takes natural language inputs and transforms them into internationally patented data structures, which it calls canonicals. Once language is encoded in this form, Mind AI says its engine can make connections, perform logical  reasoning and generate intelligent responses, without the need for massive amounts of data and computing power. The company calls it a new form of AI. Differentiators include transparency (i.e. ability to trace-back how certain conclusion was made with precision), a dramatically higher rate of accuracy ( relative to LLM based systems), context hopping and its ability to handle more diverse languages natively than other types of AI systems.

The company, formed by two serial entrepreneurs, has attracted strategic investments from early-stage VCs and family offices of major business conglomerates and high net worth individuals in South Korea, the Philippines, Thailand, and Canada. In addition to those countries Mind AI has a presence in India with plans to target Europe and North America.

The mission is to translate intelligence into a machine and build the most advanced natural language reasoning AI. “We want to be the CPU [Central Processing Unit] of AI – the reasoning algorithm that anyone can use,” says co-founder and CEO Paul Lee M.D., a serial entrepreneur and a medical doctor by training.  “What we can do that nobody else can is context hopping and transfer of knowledge,” says Lee. “If you understand the logic in buying things or booking hotels you should be able to apply this knowledge to booking airplane tickets or even buying things in supermarkets.”

His co-founder, Joshua Hong, a technology entrepreneur, financier, and web3 architect, previously worked at Accenture as a technology consultant and Deutsche Bank in its global investment banking division. He then started North America’s first free-to-play MMORPG (massively multiplayer online role playing game) publisher, GamersFirst, which became one of the largest free-to-play MMORPG publishers in the Western Hemisphere with offices in U.S, India, Brazil and Turkey. Hong also started a virtual game asset trading platform in China, Item*Star, while simultaneously helping to build Playspan, the largest virtual currency payment platform for gamers in North America. He acted as Playspan’s early backer, shareholder, and board member prior to its acquisition by Visa International for $190 million in 2011.

Both Hong and Lee also co-founded Synesis One, an AI data crowdsourcing platform on the Solana Blockchain, and Curely and Kuddly, growing that platform into one of the first mobile telehealth platforms in the U.S. for the health and pet health industry, with close collaboration with IBM Watson.

The two co-founders have spent the last 13 years working on natural reasoning technology, much of it in stealth mode.  Mind AI was launched in 2018. The S. Korean startup closed a $7 million seed round in 2021 and started commercializing the technology in April of this year. Lee says Mind AI has completed proof-of-concept trials with a major car manufacturer as well as healthcare companies, gas utilities, supermarket chains and e-commerce companies. The company says it has contracts with 17 clients, with 60 more in the proof-of-concept phase.

Customers include Villa Market, a Thai supermarket chain, Undenna Group, a conglomerate in the Philipines, Yesco, a city gas supplying company owned by South Korea’s LS Group; Share Investor, an investor relations company owned by the Thai Stock Market, and Kyowon Group, a South Korean education conglomerate

LLMs’ Limitations

The most common approach to AI today is Deep Learning, which requires enormous amounts of computing power and data to train algorithms. There are data privacy issues as LLMs use their clients’ data to train their engines. Among the limitations of the deep learning approach is that inference is done through pattern matching with no understanding of context. No reasoning is possible, and the process is a black box.  ‘Knowledge’ gained is generally not applicable to any other domain, says Lee. What’s more LLMs don’t have any understanding of the world and the algorithms hallucinate, giving wrong or inappropriate answers with confidence. Lee compares it to trusting “a drunken uncle at a party.”  Business needs other options, he says.

A Different Approach

The challenge for Mind AI was to get their AI engine to comprehend and reason like humans do and infer things naturally by understanding the meaning, not just guessing what is going to come up next. “We educate our models, not train,” says Lee. The startup’s secret sauce is the ontologies it has built, which create semantic relationships between words, phrases, and sentences, he says. “It is like educating a little child not only about definitions but cause and effect,” says Lee. “To achieve this, we collect different kinds of data about how the world works. That is what we call common sense knowledge, and it is different in every culture and country. That is why we use a canonical model so that the network can consistently evolve with new information. “

Mind AI uses this engine as the core of different product offerings, including conversational AI which it describes as “chatbots on steroids.” Its conversational AI product can do context hopping – ie – change topics – and because the service is based on a canonical network can explain what sort of thesis it used to come up with an answer and if the answer is wrong, go back and correct it in real time, says Lee.

The South Korean startup says its chatbots excel at natural conversation in multiple languages, not just in English and Chinese. “There are other cultures out there, says Hong. “By supporting languages like Korean, Thai or Arabic, we are ensuring there will be diversity.”

Mind AI’s “human logic intelligence” combined with computational AI is also being used for robotic process automation and applied to inventory control, supply chain management, credit card processing, insurance underwriting and more, he says.

Villa Market, a Southeast Asia online and offline grocery chain, has applied Mind AI’s technology to several use cases, says Lee. It had a large volume of products that needed to be checked daily and completed all these processes manually, a time-consuming process. The project goal was to provide an AI that can provide logical communication with machines to identify faulty products, pinpoint root causes and mitigate potential business failures. Mind AI says it was able to achieve a 90% automate rate and 100% accuracy rate.

Separately, Villa Market, asked Mind AI to build a conversational AI. It was using Google’s Dialogue Flow product, which took eight months to build and had an accuracy rate of only 82%, according to Lee. “We launched our product in 10 days -which it is still using – and it had an accuracy rate of 96%,” Lee says.

Mind AI says its technology can also be used to transform digital documents into canonical repositories that can give answers to queries about the subject from diverse sources.

An Increasingly Competitive Field

Of course, Mind AI is not the only company working on alternative approaches to AI.

China now has at least 130 LLMs, accounting for 40% of the global total and just behind the United States’ 50% share, according to brokerage CLSA.  Additionally, Chinese companies have also announced dozens of “industry-specific LLMs” that link to their core model.

As uptake of Chinese and U.S. LLM models takes off other countries risk becoming increasingly dependent on foreign AI models.

Startups such as France’s LightOn and Germany’s Aleph Alpha are trying to develop alternative LLMs that safeguard data. So is Open GPT-X, an initiative by ten German organizations from business, science and media that is developing the European answer to GPT-3, that includes the Center for Information Services and High Performance Computing (ZIH) at TU Dresden. The German Federal Ministry for Economic Affairs and Climate Action is funding the Open GPT-X project within the Gaia-X funding inititative with around €15 million. Under the leadership of the Fraunhofer Institutes for Intelligent Analysis and Information Systems (IAIS) and for Integrated Circuits (IIS), the OpenGPT-X project goal is aiming to develop a large AI language model for Europe that offers data protection as well as European language diversity.

Israel-based start-up AI21, while less well-known than its rival OpenAI, is a serious challenger in the market. Users of AI21 studio can train their own versions of the LLM with as few as 50-100 training examples, which then become available for exclusive use. AI21 also offers WordTune Spices chatbot, which distinguishes itself as a ChatGPT alternative by the use of live data retrieval and the citation of sources in its formulations.

Founded by former OpenAI employees U.S.-based Anthropic has launched its own large language model, Claude, a ChatGPT alternative based around what it calls “constitutional AI”. The model is designed to act according to programmed principles (i.e. its ‘constitution’) Google has invested $300 million into Anthropic for a 10% stake in the company.

Elemental Cognition, another U.S. company, has developed a hybrid AI platform that combines natural language understanding, machine learning, explicit knowledge models and automated reasoning to enable a new class of AI applications capable of learning and delivering explainable intelligence.

Mind AI’s says its biggest competitors are OpenAI, Google, Microsoft and Baidu, the companies behind the LLMs with the most traction to date.

“We know we can’t outrun Big Tech by just tweaking something, so we set out to build something completely different, and do it independently, while ensuring that in the end it would complement and work with LLMs,” says Lee.

This article is content that would normally only be available to subscribers. Sign up for a four-week free trial to see what you have been missing.

To access more of The Innovator’s Startup Of The Week articles click here.




About the author

Jennifer L. Schenker

Jennifer L. Schenker, an award-winning journalist, has been covering the global tech industry from Europe since 1985, working full-time, at various points in her career for the Wall Street Journal Europe, Time Magazine, International Herald Tribune, Red Herring and BusinessWeek. She is currently the editor-in-chief of The Innovator, an English-language global publication about the digital transformation of business. Jennifer was voted one of the 50 most inspiring women in technology in Europe in 2015 and 2016 and was named by Forbes Magazine in 2018 as one of the 30 women leaders disrupting tech in France. She has been a World Economic Forum Tech Pioneers judge for 20 years. She lives in Paris and has dual U.S. and French citizenship.