Chatbots, Thought Leadership

Chatbot Unlocked: A Q&A Series With Adi Gaskell

Adi is an experienced writer on technology, startups, and innovation, with a special focus on artificial intelligence. He has also delivered talks on the subject for the likes of the NUJ, the Guardian, Stevenage Bioscience, EdTechX, the House of Lords, and CMI, whilst also appearing on shows such as BBC Radio 5 Live and Calgary […]

By Adi Gaskell
December 29, 2020

Adi is an experienced writer on technology, startups, and innovation, with a special focus on artificial intelligence. He has also delivered talks on the subject for the likes of the NUJ, the Guardian, Stevenage Bioscience, EdTechX, the House of Lords, and CMI, whilst also appearing on shows such as BBC Radio 5 Live and Calgary Today.

What is your background and/or expertise in this space?

Academically I studied computing with artificial intelligence and developed a chatbot during that course (back in 2003). This experience then led to the development of a chatbot for the website of a non-profit venture I created after graduating. Suffice to say it wasn’t as sophisticated as modern chatbots, but I’ve maintained an interest in the space ever since.

How do you work with chatbots and artificial intelligence in your current role/industry?

My current role revolves around the discovery of interesting new technologies, whether that’s in a consulting, advisory, or journalistic role. Obviously, AI plays a fundamental role in many of these new technologies, and chatbots are one of the more interesting applications of it. I have a particular interest in healthcare and med-tech, and so it’s been interesting to see chatbots emerge, and be useful, in areas such as mental health.

How has chatbot technology and artificial intelligence adapted, and in some ways accelerated, amid the Covid-19 pandemic?

Obviously, the pandemic has created a forced distance between us, while also significantly increasing the demand for digital services as we strive to retain some kind of normality. The use of AI to help meet that demand has been vital. Recent analyses of the various digital investments made during the pandemic do, however, reveal the way we still have to go for AI to become more mainstream, as most investments have been in areas such as remote working/business continuity and cybersecurity. AI remains a relative minority activity, despite those who do invest achieving so much with it, whether in terms of productivity, revenue or profitability.

What are some best use cases of conversational AI and why?

I think the applications in mental health have been the most interesting for me. It’s been shown that having virtual interfaces can help people to open up on extremely sensitive topics, whether that’s post-traumatic stress disorder or eating disorders. The virtual platform provides a distance that reduces inhibitions, which has been a really interesting finding as traditionally digital has been seen as lacking the human touch.

What are some unexpected uses of artificial intelligence users might encounter in their everyday lives?

I think one of the most interesting areas is in natural language processing, especially in translating languages. I’ve been learning Czech for a few years now due to a Czech wife, and I’m losing the race to master the language before AI is able to translate both written and voice commands effectively. Czech is notorious for its difficult, and AI-based services are already pretty good at translating it into English (and vice versa). The ability to communicate in our global world is so important, especially in areas such as healthcare and education, so I think this will be a significant breakthrough in the coming years.

What do you think is the biggest obstacle bot developers face when trying to create a conversational bot? What are some ways to overcome these obstacles?

Languages are rich and nuanced things, so understanding slang, idioms, and even jargon is fundamental to how we communicate but very difficult for machines to comprehend. As our ability to mine ever larger datasets improves, however, our ability to understand the various nuances of language is growing too.

What is the biggest pain point for chatbot developers or chatbot users today?

I think it’s important to understand that chatbots are part of a wider process, whether that’s delivering customer service, financial advice, or mental health support. There will be times when a chatbot is adequate for our needs, and there will be times when, for whatever reason, it isn’t. Too often companies produce a chatbot and neglect other forms of service, which can leave users frustrated if they don’t get the answers they need.

What technology or functionality is missing from chatbots that you think will help with consumer engagement or adoption?

The ability to overcome problems in human behavior still exists as a lot of the data that is being fed to AI remains as biased as humans inevitably are. We need AI to do better than humans, so still need to work on eradicating those biases.

A recent poll to our Facebook audience revealed that they find the biggest drawback of conversational AI to be users’ fear of its capabilities in the future, such as automation overtaking jobs. What solutions to this concern would you propose?

I actually wrote about a recent study from Stanford that showed that companies that invest most in AI are actually creating more jobs than those that don’t, so I feel that while there will be an obvious need to adapt our skillsets to the changing world, there won’t be the widespread jobs apocalypse many fear.

Our audience also prefers to develop and interact with more voice bots in the future. Why do you think this preference exists, and what advice do you have for new developers?

The rise in services such as Siri and Alexa undoubtedly plays a major role in that, so it is perhaps not all that surprising. It would be recommended for any chatbot developer to work on both text and voice-based versions to future proof themselves.

What is the “next big thing” in making artificial intelligence more human-like?

I would argue that we should be worried less about making AI more human, and more concerned with making it more human friendly. AI shouldn’t be developed to replace humans, but rather to augment what we do, so the interface between us and the technology should be a real area of focus so that our complementary strengths smoothly intersect.

What conversational AI trend(s) do you expect will take over the industry in the next year? The next five years?

I think the biggest trend will be the ability for the bots to learn for themselves rather than be fed data from the developers. We’ve obviously had some missteps in this direction with Tay and others like her, but that’s undoubtedly the next area, especially if it can be done in a way that doesn’t exacerbate some of the less salubrious parts of human nature, such as racism and discrimination.

Is it important to have a wide array of tools and platforms available for conversational AI development at this stage of the market? Or do you think focusing all our efforts on just one or two platforms is better? Would interoperability across development tools, libraries, and platforms help the market grow?

I think interoperability between tools and platforms is generally a better approach for technology generally and society doesn't tend to benefit from a walled garden approach. The market for tools and platforms is seemingly becoming semi-consolidated, however, so it is perhaps easier to identify those that will come to dominate, especially if you're targeting a specific niche.