In 1942, before the first robot had even been built, science fiction writer Isaac Asimov came up with the Three Laws of Robotics, a set of rules to govern their usage in an ethical manner. First, a robot may not injure a human being or allow one to incur harm. Second, a robot must obey human orders except when they would cause harm to humans. Third, a robot must protect its own existence unless doing so means disobeying one of the first two laws.
More than 75 years later, and 64 years after the first robot was developed, Spanish startup Acuilae has developed a new AI module, Ethyka, which applies ethical and moral standards in complex AI systems. Combining machine learning, deep learning, data analysis, statistical models and advanced data visualization, Ethyka is able to analyze a recognized dilemma in several phases.
First, it determines the type or types of ethics to be used, e.g., civic or professional. Second, it assesses the principles to be met depending on the context and the type(s) of ethics to be used. Finally, it is able to generate predictions based on historical information to guide decision making.
CompassList interviewed CEO Cristina Sánchez, a data scientist who spent almost 20 years working with large clients at Accenture before co-founding Acuilae and creating Ethyka, the company's first product.
This interview has been condensed and edited for clarity.
What inspired you to build Acuilae and Ethyka?
Acuilae is a company that specializes in AI. From experience, we realized the importance of applying ethics to the mathematical algorithms that are part of AI. We are living through a new industrial revolution, and there is a certain fear in society about the lack of control we may have over AI, as reflected in films like Terminator. The Ethyka module enables citizens to see robots or artificial intelligence as tools that are here to help man, not to replace him. They do not come to take jobs; they come to change them.
When developing chatbots, virtual assistants or other tools that work with AI, these tools are enriched by everything they learn through their different interactions. They learn the good and the bad. The Ethyka module was born from the idea of giving AI a more cordial vision.
If, for example, we are talking about decisions about a supply of stock. Generally, no ethical component is needed. But what happens in the case of medical diagnosis, when a patient’s life is at stake? If there is a supply shortage, how do you determine who does and does not get treatment? It is important to remember that computer science and AI must help men make decisions, not make them for us.
What exactly is Ethyka?
Ethyka is an AI platform safety system that, due to its originality, has had great impact. It is a module to solve ethical dilemmas to be implemented in virtual assistants, autonomous cars, autonomous units, apps, web services...
Thanks to Ethyka, we can condition the learning of AI tools using an ethical component. Thus, we imitate the way the brain, which has an ethical component, thinks, and we can transmit to society that AI is being used correctly.
The idea of Ethyka arose precisely when we appreciated the existence of this gap in the AI world. It is important to create AI, but it is essential to give it an ethical component. A module with the information obtained by a machine learning algorithm, through the application of a prediction algorithm, analyzes or weighs dilemmas in light of ethical principles and historical data, before proposing the adoption of a certain decision. Of course, the adequate response of the predictive algorithm also requires training until data are assimilated and the machine knows how to use the data properly.
We must also bear in mind that ethical principles are not universal, but depend on the society in which they operate. Ethyka can take into account different types of ethics and different ethical principles, such as those present in a country, a sector or a company. The Ethyka module is designed to help decision making, increasing the information taken into account when deciding. Obviously, a human must have the last word.
How does Ethyka work? Or how would you explain how the product works to a layman?
Someone who does not understand anything about AI probably will not understand. I think the most important thing is to grasp the general idea of the importance of including an ethical component within AI.
Can you describe the three levels of the platform?
Ethyka helps to avoid AI corruption, helps AI solutions to adopt better decisions with a human component. People learn ethics or what is right and wrong throughout their entire life. In the case of AI, it results from conditioning and training. Ethyka is a three-level platform that can recognize 90% of ethical dilemmas.
To develop Ethyka, we studied the medical-scientific procedures used in cases in which a person, as the result of an accident or illness, is not able to distinguish between right and wrong. We also studied the processes that help these people recover, and we discovered that, in all cases, the processes follow a structure: dilemma, analysis, decision.
The idea of dividing the Ethyka platform into three levels came from Pedro Diezma, my partner, although I built the module. The module implements the ethics component [dilemma, analysis, decision] in the machine based on studies of how the brain works and reactions related to ethics. Of course, to give it life, a structure composed of different database, files, etc., was needed.
Can you give us some examples of dilemmas? And of Ethyka's applications?
By dilemma, we mean problems that can be solved through more than one solution, none of which are completely acceptable or, the opposite, all of which are acceptable. A dilemma leaves the decision-maker in doubt, debating between different alternatives. Ethical dilemmas can be recognized in many different scenarios, from thermographic data captured with drones to data analysis for the optimization of sports performance to data received by a virtual voice assistant for the elderly that aims to solve problems of loneliness to crowdfunding data analysis.
Ethyka can be applied to cars, robots, virtual assistants, chat boxes, autonomous units or simply any application or web service. In the logistics sector, Ethyka can be incorporated into image recognition systems to detect anomalous situations such as accidents, medical problems or a shortage of materials, or inappropriate behavior, like theft, and can alert people. In the financial sector, Ethyka can predict anything from default risks to anomalies in operations such as fraud.
Do you have some kind of criteria to establish the basic ethical principles?
The basic ethical principles are those established in the corresponding society. In this case, in Spanish society. If they ask us for an ethical module for another country, it would be necessary to study the basic ethical principles that correspond to that country. The ethical principles are reflected in the module through a series of combinations or conditions implemented within it.
A classic example to understand why it is necessary to include an ethical component in AI is to imagine what would happen if we had a chatbot that we are training to interact with the elderly and took it to a high school to interact with teenagers. The chatbot would be corrupted in a couple of days, similar to what happened with Tay, the conversational robot created by Microsoft for Twitter, that had to be removed when it began to issue inappropriate comments after a few hours of interacting with Tweeters. With an ethical component, however, the chatbot can block offensive language and not learn from what it sanctions as unethical.
What applications or technologies do you use?
We are compatible with any technology. We use Open Source technology, data science, and AI technology such as IBM Watson or Microsoft's. If another company uses another type of AI technology, the module is fully compatible with it. API and web services, security technology and HW installing firmware are also used.
Could you tell us about the background of the founders and the team?
Our whole team has extensive experience in their fields. My partner and co-founder of Acuilae, Pedro Diezma, has a degree in Business Administration and Management, is a professional speaker and a top IoT and wearables influencer, according to Onalytica. He is considered the best worldwide in terms of smart vest devices, smart watches, etc. He is also the founder of Zerintia.
I am a statistician, a computer scientist by profession and a data scientist. Thanks to my work in a multinational company for almost 20 years at Accenture Spain, from 1998 to 2017, I can now lead my own AI company and have credibility, since I have a lot of experience behind me.
At Acuilae, Pedro Diezma is the visionary, and I am the technician. We have very different and complementary profiles. Pedro is dedicated to marketing, public speaking, business plans and carries out the financial part. That is, the horizontal aspects of the business. I take care of the technical part, the AI, clients and client proposals and dealings with suppliers.
What milestones, difficulties and successes from your journey as a company do you consider most important?
Ethyka first appeared in April 2018, and, since then, the company has been featured in numerous contests including South Summit Madrid 2018. We consider being one of 100 finalists out of 3,500 companies from 84 countries at South Summit an important achievement.
Key to our success have been the different skillsets Pedro and I bring to the company. I know that I am a very good technician, but I do not know what doors to knock on on a company level. It is clear to me that alone I could not have achieved what Pedro and I have done together.
Regarding difficulties, if you want to scale production, financial support is needed.
How has the company been financed?
We have been bootstrapped to date, but we are looking at a financing round. With €300,000 we would be able to produce and market our MVP.
From then on we would be implementing a lot of other functions. But it is a very complex module. We cannot do something very complex in a short time and implement a lot of functions. It would not be realistic. We have to go step-by-step and respect the timelines. Otherwise, we would be risking all the time, effort and work that we have dedicated to the Ethyka prototype and our company, which does not make sense.
What type of companies or sectors are you targeting?
We have a range of people or companies with whom we are collaborating or in contact. For example, there is a large marketing company that wants to collaborate with us on a project of considerable importance. Next week, we have an important meeting with a robotics company. We are in contact with digital companies, banks and financial institutions, logistics and transport companies, health and healthcare companies and car manufacturers.
For now, we are looking for collaboration agreements with customers or companies, in addition to promoting our company and our product. We are still perfecting the prototype. It is not yet commercialized.
We receive many requests to attend congresses, both national and international, and to give lectures and seminars. For example, the other day we received a call from Geneva to attend an ethics conference.
As for partners, at the moment Amazon is talking to us, Microsoft is talking to us and IBM was talking to us yesterday. We are in the phase of getting agreements.
What is your business model?
We have thought about three different lines of business. On the one hand, to develop modules on request, adapting the module to the specific needs of the client. In this case, we also do maintenance and training.
Another line of business would be oriented to telecommunications companies. In this case, the structural module would be sold, as a skeleton, to be adapted by the client’s developers.
The third line would be a SaaS or payment for use service in the Cloud, using our platform. Thus, no customization is possible. In this case, the devices or AI tools call to connect with our Ethyka module through the Cloud.
We are targeting the Spanish market initially. But we are also aware that Ethyka is a module that can be marketed internationally because it is a function that is necessary for all markets and countries.
Who are your competitors?
In Spain, we have no competitors for Ethyka, although we do as an AI company. In this sense, there are large consultancies with their own AI departments and more startups that provide AI services.
Worldwide, we are not aware of any competitors for Ethyka, although our press releases have had a lot of repercussions at the international level. There are other AI companies around the planet.
What would you like to see in the development of AI in Spain?
On the one hand, people today still do not know what AI really is. On the other, AI companies are growing in number and size. Therefore, how far anybody goes in this field depends on their personal ambition. There are still a lot of young people being trained. The future depends on a desire to learn, but also on the need to build trust.
Companies want to implement AI, but professionals must explain what is possible and what cannot be done. To make prototypes and get investment to market them is great, but they need to work when implemented. Algorithms always have a failure rate, which is sought to be the minimum possible. But if some algorithm is 100% successful, it is simply flawed. That's the really important thing, to understand that, whether it's AI or a machine, it must be ultimately controlled by a human. We do not have to believe for sure what an algorithm is telling us.
When anyone comes to ask us about Ethyka, I tell them we are in the prototype phase. To implement it and have it working in a consolidated company, I need an investment to be able to create a great set of training data, so that when I implement it, I will be sure that it fails at a very low rate.
How do you see the AI market in the future?
AI is something that is coming upon us, like a tsunami. However, there are very few people trained to work in AI. In fact, it might be a problem for us in the future, when we start to grow. On the one hand, there are large consultants and, on the other, small companies. The people who are prepared in this area are highly valued and forecasts are that it will continue that way. For example, in the US, a data scientist earns a lot of money. In Spain, however, not so much.
Companies want to implement AI to avoid a competitor doing it before them, because, even if it means an investment, AI will reduce costs in the long run. With AI, offering solutions to customers will be less expensive.
However, we must not forget that algorithms have a percentage of accuracy or credibility, and a percentage of non-credibility. That is why there must always be a human who makes the final decision.
