ECJ EUROPEAN Cleaning Journal

Artificial intelligence has transformed our lives in ways that were the stuff of science fiction not so long ago and its influences are set only to proliferate over coming decades. Hartley Milner examines the impact of AI on the business world, for better or worse.

How would you feel about interacting face-to-face with a computer-generated ‘human’? I mean a fully self-learning AI avatar that can hold lucid conversations, give informed answers under close interrogation and make intelligent and creative decisions, all while maintaining eye-contact and responding to your facial expressions, gestures and even emotions.

Such super-smart humanoids are being touted as the next big leap in the evolution of AI. And the technology is advancing so rapidly that by 2050 small businesses in the UK will be run by personal avatars, according to digital bank Virgin Money. Avatars will be programmed by SME leaders using their company’s persona and be authorised to carry out functions and make decisions on their behalf, leaving them free to focus on the most important tasks.

One reason for the rise of AI avatars in use today is that people say they prefer seeing a human face when interacting with a business or making a purchase. Because more sales now take place online than at retail locations, the technology gives transactions a human face that a website or app cannot replicate. As avatars become evermore sophisticated, you will find yourself engaging with them as naturally as you would a human colleague, their developers claim.

So what makes these digital clones so damn smart? Simply put, avatars learn through deep-learning algorithms programmed in to make them more proficient each time they engage with a user, mimicking how we ourselves learn, adapt and behave in certain situations. Their ability to engage in conversations with people is derived from using natural language processing (NLP) algorithms. AI avatars are a type of chatbot, but more advanced in terms of interaction.

But perhaps they are not quite as smart just yet as the hype suggests. “AI technology has made amazing strides in recent years, but at this stage of development its promise does not always deliver the reality,” Miranda Pearman, a tutor at the Birmingham Institute of Technology, told ECJ. “For one thing, most avatars still visually resemble cartoons more than real humans and lack the animation in their modelling that would make their movements and gestures totally convincing.

“This may be fine for some applications of the technology, like in marketing, ecommerce, customer service and training. But in our face-to-face interactions with real people we consciously or subconsciously pick up signals from their facial expressions – little things like a slight raise of the eyebrows or the briefest hint of a smile – that help us determine if we can trust or believe what they are saying or thinking.

Subtle responses

“Because virtual humans lack the ability to think in the way we do we don’t get these subtle, spontaneous responses in our interactions with them. And yet nonverbal signals make up a
huge part of daily communications. In fact, body language may account for between 60 to 65 per cent of human communications. More importantly, however brilliant virtual humans may be at processing data quickly, they currently lack the cognitive ability to engage with us on our intellectual level, which could make having reactive conversations with them difficult and frustrating.”

And creating humanlike AI able to process information or think like humans will be crucial if the technology can ever to be fully relied upon, according to research by the University of Glasgow’s School of Psychology and Neuroscience. The study team used 3D modelling to analyse the way deep neural networks – part of the broader family of machine learning – are able to process information to learn how their processing matches that of humans.

Professor Philippe Schyns, who heads up the university’s Institute of Neuroscience and Technology, said: “When building AI models that behave like humans, for instance to recognise a person’s face whenever they see it as a human would do, we have to make sure that the AI model uses the same information from the face as another human would do to recognise it. If the AI doesn’t do this, we could have the illusion that the system works just like humans do but then find it gets things wrong in some new or untested circumstances.”

It is hoped the research will lead to the creation of more dependable AI technologies that behave more like us and make fewer unpredictable errors, and when they do drop a clanger make it easier to understand why. AI tech developers are already racing to be the first to crack the ‘thinking machines’ conundrum. Meanwhile, tech corporation Meta (formerly Facebook) is investing in software capable of learning about the musculoskeletal system of the human body to make avatars more relatable for us.

ChatGPT

Another piece of AI wizardry that has got us all talking about it of late is ChatGPT… a generative AI language model that can create various types of content, including text, imagery, audio and synthetic data. You may recall that its emergence late last year triggered alarm among educationists that students could use the software to write essays or answer homework questions. Lecturers at UK universities have been urged to review the way in which their courses are assessed amid concerns that students are already using the potent new tool to write their dissertations.

However, with close scrutiny, generative AI is proving a boon to businesses, having applications across a range of functions, including:

• Marketing & sales – crafting personalised marketing, social media and technical sales content and creating assistants aligned to specific businesses, such as retail;

• Operations – generating task lists for efficient execution of a given activity;

• IT/engineering – writing, documenting and reviewing codes;

• Risk & legal – answering complex questions from vast amounts of legal documentation, and drafting and reviewing annual reports;

• Research & development – accelerating drugs development through better understanding of diseases and the discovery of chemical structures.

Trust issues

Now developer OpenAI has launched an upgraded version, GPT-4, which the company claims “is more reliable, creative and able to handle much more nuanced instructions”. Plus it can recognise images and provide answers in text, and has a sense of ethics more firmly built into the system than the old version. “GPT-4 is 82 per cent less likely to respond to requests for disallowed content and 40 per cent more likely to produce factual responses,” OpenAI says.

But it cautions: “Despite its capabilities, GPT-4 has similar limitations to earlier GPT models: it is not fully reliable, has a limited context window and does not learn from experience.”

Issues of reliability around artificial intelligence, and therefore trust, are a concern for the European Union, which is legislating for how it is used. The EU AI Act – expected to be in force by the end of this year – will regulate high-risk AI use in areas such as law enforcement, infrastructure, product safety, justice and surveillance, and where it could be used to generate ‘deepfake’ content or throw up cyber security threats.

Today, artificial intelligence is proving its worth in a supporting role rather than as a substitute for human intelligence and ingenuity. AI currently has difficulty carrying out common sense tasks in the real world. Its big strength is that it can analyse and process troves of data many times faster than the human brain and then come back with synthesised courses of action. In this way, the user has the information they need to look at various possible scenarios for dealing with a particular situation and make better-informed decisions.

But what of tomorrow? Experts have identified the top five job categories that are most likely to be replaced by AI:

• Technical – programmer, software engineer, data analyst;
• Media – advertising, content
creation, technical writing, journalism, graphic design;
• Data – accountants, traders, marketing research analysts;
• Financial – financial analyst, personal financial adviser;
• Language – customer service, teachers, lawyers.

Miranda Pearman’s last thoughts on AI: “While the technology is good at some jobs it is currently not smart enough to fully replicate the range of skillsets needed to do most types of work. For example, despite the billions spent on developing self-driving vehicles we have yet to find ourselves making journeys in buses, taxis and trains that do not have a human at the controls.

The technology is simply not trustworthy enough, and the same situation exists across the vast majority of employment sectors, especially where there is an emphasis on human-to-human interaction, like nursing or teaching.

“Perhaps think of AI as an adolescent child transitioning into secondary education. It is a time when the child gains more advanced capabilities and widen their circle of influences, and learns to think for themselves. It is also the time when they are most open to the influences of educators and others, which helps shape how they turn out in adulthood. It will be down to humans how AI develops and impacts on businesses and therefore on jobs… and, indeed, across every aspect of our lives.”