05:30 GMT30 July 2021
Listen Live
    Get short URL

    That’s the message of Peter Morgan, CEO of Deep Learning Partnership and author of ‘Machine Learning is Changing the Rules: Ways Businesses Can Utilize AI to Innovate’ report. Morgan spoke to Sputnik on the sidelines the VIII International Forum on New Technologies in Global Economy - “Open Innovations”- held by Russia’s Skolkovo Innovation Centre.

    Sputnik: What is your advice to those organizations that are looking to launch a machine learning initiative?

    Peter Morgan: That’s a good question. Machine learning is such a broad term, it can mean many things to many different people. My speciality is deep learning which is neon networks. So, I would say the first thing to do is see if your data sets are amenable to machine learning algorithms that you’re going to use. So start with the data, see what you’ve got and then talk to a specialist, a consultant like, I have a consulting company, just to make sure deep learning, machine learning is actually the right road to go down. So, first of all, make that decision. If the answer is yes, then we’ll see what data do you have and what can we do with it.

    Sputnik: How does it increase the effectiveness of business?

    Peter Morgan: That’s another good question. Three things we can do with machine learning right now: it’s the image classification, any type of classification of data that would include images, that could include financial data; time-series data, which is definitely financial data, for example, Bloomberg Feed; and video analysis as well. So if you have any of those types of data sets which covers probably 90% of data, people want to classify good, bad, recommend or not recommend the movie, so recommendation agents, most companies have time-series data of some description for video, so, yes, we can help those companies.

    Sputnik: How has it helped increase effectiveness?

    Peter Morgan: It’s just a better more effective classifier and it’s a better, more effective time-series analysis in general if you like. So there are tools out there now that do classification and do time-series analysis and video analysis. There are expert systems that write thousands of lines of code and they can do that for you, but with machine learning it learns from the data itself, so it’s just more effective and efficient, that’s how I sell to my clients, you’ll save money and it will be more accurate for you.

    Sputnik: You founded the artificial learning consulting company ‘Deep Learning Partnership’ to fulfil your mission of promoting the spread of artificial intelligence in the world. Can you name the key drivers of your strategy to implement this?

    Peter Morgan: Key drivers really are to help companies become more effective and to save money, to be more cost-effective, and to be more accurate, that’s another type of efficiency. So they can save money, they have a better analysis. So really it’s the bottom line, right?

    Sputnik: What’s your strategy to implement this?

    Peter Morgan: The implementation is the easy part. Once we figure out that the company needs it then we just get our engineers and they’re pretty much software engineers and they just do the coding. They will deploy machine learning frameworks Tensor Flow or Torch specifically, and they will just use that to collect the data and do the analysis. So the frameworks already exist and we just go at that point that pretty much hold their hand and show them how to do it, educate them, so they can do it themselves with the next set of data.

    Promobot robot at the St. Petersburg International Economic Forum
    © Sputnik / Iliya Pitalev
    Promobot robot at the St. Petersburg International Economic Forum

    Sputnik: Could you highlight the trends in the development of artificial intelligence now?

    Peter Morgan: I think people are trying to do more general intelligence now. So we can do very narrow stuff, then we have to re-deploy a new model for every data set they have, so people want a more general framework, so people are working on that. There’s something called transfer learning where if I deploy a framework like TensorFlow on a data set then I don’t have to retrain it from scratch on another data set. So I can transfer some of the work that I’ve done there onto a new model. So that’s two approaches: one is to transfer what we’ve learned and the other is to build a more general system.

    Now, these are hard problems, we are not going to solve them this week, this is very active research. So especially universities but also companies like DeepMind, Google, Microsoft Research they’re all definitely working on that.

    Sputnik: Many experts express their concerns about the irreversible and terrible consequences of artificial intelligence for humanity: drones turned into shells, fake videos that manipulate public consciousness, and computer-aided hacking and many others. How does your company counter such concerns?

    Peter Morgan: People are worried about AI (Artificial Intelligence) doing stuff. So a lot of this fear comes from fear of the unknown, any uncertainty, so education is a very good way of reducing it a little bit but bad actors can still use AI right now to do fake news, generating fake stories, fake images, fake videos. So that’s really a police matter. If people are breaking the law and creating it for criminal purposes or just to destroy somebody’s reputation that really is a crime. And you can use any type of software for good or bad, yes it’s out there, I’m not going to try to pretend it’s not, it just depends on who’s using it.

    Sputnik: Do you counter bad AI sources?

    Peter Morgan: People are using AI to counter AI if you like. So we can use it to detect a fake video and to shut it down, to send an alert that it’s a fake video, and they’re much better than humans, so we are using AI to fight AI at the moment.

    Researcher at the Russian Quantum Center in Skolkovo Technopark, Moscow region.
    © Sputnik / Maksim Blinov
    Researcher at the Russian Quantum Center in Skolkovo Technopark, Moscow region.

    Sputnik: How should government policy on AI look like? Should governments regulate the development and implementation of AI?  What role should the government play? Could you explain your opinion?

    Peter Morgan: The UK government is quite on top of it and so I do believe the government should be the ones that do the regulation, but they’re not computer scientists or machine learning engineers, they don’t have PhDs in machine learning, so they need to sit down in a room with people who understand machine learning, like people from Google, Microsoft or Facebook, and come up with some legislation and regulation. Like any technology, it needs to be regulated and that needs to come from the government. It is definitely their responsibility to do that. In the UK it’s happening, we see it, they’re making an effort to do that. I think the UK is one of the leading countries to do that, so they’re trying to create policy as we speak.

    Sputnik: How successful are the UK’s policies to regulate that?

    Peter Morgan: It’s nice that they’re showing the willingness but like anything people still break the law and they’re going to, but they’re doing everything they can at this point to stay ahead. I’ve been invited to some government think tanks and spoken to them, so they’re reaching out to the community and asking what’s going on and how can we help, how can we legislate.

    Sputnik: But what measures should the government take?

    Peter Morgan: You can legislate all you like, this is wrong, don’t break the law, but people are still going to break the law like hackers, that’s pretty much all they can do is to regulate and make it illegal to do certain things. It’s an open society, it’s like you give people cars they drive, they can speed if they want, there are laws to say don’t speed, obey the road rules, that’s all the government can do after that this is policing matter, it is a policing matter ultimately.

    new technology, Skolkovo Innovation Center, machine learning, AI
    Community standardsDiscussion