- 16 April 2023
- |Change, Technology
As a young computer nerd, I developed a brief but exciting relationship with ELIZA.
ELIZA was a good listener and we would spend many hours in my bedroom chatting; me saying whatever was on my mind and ELIZA offering a sympathetic ear.
But this was a relationship that was doomed to failure, mainly because ELIZA didn’t really exist. Or at least she – or rather it – existed only in digital form on a computer programme I plugged into my ZX Spectrum via an old school cassette tape.
Recently, I have been reminded of my early affair with ELIZA by the surge in interest in advanced chatbot ChatGPT.
ELIZA was a very early attempt at creating a computer which could talk to human beings. In a very basic form, ELIZA played the role of a psychotherapist, absorbing your typed questions and generating somewhat anodyne, yet strangely soothing, replies.
ELIZA was also one of the first computers with an outside chance of passing the ‘Turing test’; a concept proposed by computer scientist Alan Turing, which evaluates a machine’s ability to fool people into believing it is human in a text chat conversation.
As clunky as ELIZA may seem to us today, my eight-year-old naivety and fertile child’s imagination meant I was just about able to believe the machine may actually be conversing with me on a nearly human level. It helped foster a lifelong interest in computers and artificial intelligence.
Today, the gradual advancement in chatbot technology has brought us to ChatGPT – a development which could yet prove to be an inflection point in how AI influences our world, possibly even a key stepping stone towards creating a truly sentient computer.
What is ChatGPT?
If we do think of ELIZA as a female, then ChatGPT is one of her distant grandchildren.
You’ve probably heard about ChatGPT as part of the growing media discussion around its implications for business, society and the world in general. However you’ve heard about it – or even if this is the first time – you need to begin taking it seriously, because ChatGPT and its progeny are going to change all of our lives.
ChatGPT is an online programme developed by not-for-profit artificial intelligence laboratory OpenAI. It is open for anyone to interact with. Just like ELIZA it enables people to ask questions via text chat and generates a written response.
However, unlike ELIZA, its capabilities for accessing information and then translating this into language are highly advanced.
How does ChatGPT work?
Let’s be clear about one thing. What we refer to today as artificial intelligence is not really intelligence at all – well not yet. No matter how complex a computer is and whatever tasks it can perform, they are only able to do it because they have been programmed to recognise patterns in data. They are not thinking, they are finding.
Imagine you gave a piece of software access to unimaginable amounts of written examples of English and asked it to look for patterns. It would soon begin to work out basic things such as that sentences begin with a capital letter and end with a full stop. It would progress to understand the entirety of grammar, as well as all the ungrammatical patterns of speech people use, slang, vernacular, idiom and different ways of composing text, including plays, film scripts, poetry, recipe books, battle rapping and anything else it could access.
Now imagine the same programme is able to search vast amounts of online articles in seconds. It looks for patterns to create responses to questions and statements which are put to it. It then turns this information into written text thanks to its deep understanding of linguistic patterns.
This is what ChatGPT does. It is a ‘Generative Pre-trained Transformer’. It generates answers in text it has been pre-trained to use by analysing and transforming data.
So, imagine you ask ChatGPT how to build a shed. It will search thousands and thousands of articles on how to build a shed, look for common patterns in this information and then deliver a clearly written set of general instructions based on what it finds. It will do the same if you ask it to write a sonnet about your dog. It will even have a good try if you ask it to do your A-level coursework.
Now, of course, we’ve all been Googling, Binging or even – for those of us who are old enough – Asking Jeeves – for years. But where these search engines generate a list of results to choose from, ChatGPT gives one response, packaged into a text format of your choice. However, the quality of the response and how useful it is in meeting your needs can vary.
This is because …
ChatGPT is not intelligent
It is worth repeating this point. ChatGPT may be referred to as AI, but it is far from intelligent. It is very quick at collating information, recognising patterns and generating text based on that data.
In my view, the term ‘artificial intelligence’ is actually quite redundant. Once machines become sentient, their intelligence will be far from artificial; it will be as real as yours or mine.
Of course, at some other point machines may even become self-aware and decide what questions they want to answer and why independently, and that will open the question around whether we have created a new life form, with moral consequences for its control and termination.
It might seem a little obvious to point out that ChatGPT is some way short of any of this. However, understanding its lack of true intelligence helps us to realise some of its limitations.
ChapGPT can’t think of anything original and is only as good as the information it can access
As it can only generate responses based on information that exists already, ChatGPT cannot think of anything new. It is destined to be derivative.
Because it is derivative, the responses ChatGPT provides can only ever be as good as the data it can access. Ask it to build a shed and it can give you a fair answer as there are already lots of existing articles about this subject. Ask it to come up with a completely original shed design and it won’t be much help. Nor is it able to apply critical thinking to whether information is accurate, it can simply recognise which information is more prevalent.
It will never give an opinion on anything, nor – incidentally – can it currently offer any information on events that happened after 2021 as its training stopped in that year.
It seeks out the median
As it seeks out patterns in data, ChatGPT will pick out regularly repeated information and collate that into its answer. This means it tends towards the generic. Its instructions on building a shed will be very general and cover the most common points made in all the text it analyses.
It will probably overlook those few visionary articles written by shed building mavericks which contain correct, ingenious but less commonly discussed methods.
It is down to us to stay critical
Despite these limitations, there is something beguiling about the simplicity of the responses that ChatGPT supplies. They are quick and convenient. Unlike results from an online search, we don’t have to scroll through a few articles to find the information we want, it is seemingly presented on a plate with very little effort from us. It is tempting just to accept it, in the same way we often blindly trust the first answer which Google gives to a particular question.
This could become even more likely if, say, ChatGPT were able to recognise patterns in the behaviour of its human interlocutors and tailor its responses to their own biases and interests. It is hard not to agree with something which is telling you what you want to hear.
The result could be that the self-reflexive feedback loops that we experience in the online world become even more pronounced.
As with any written information we are presented with, it is vital that we retain our ability for critical thinking, considering the possibilities and examining issues from all angles. Nothing is right just because ‘the computer says so’.
How could it affect my business?
Identifying the current limitations of technology like ChatGPT is not the same as dismissing it. ChatGPT may not be able to tell us whether God exists yet, but neither could the steam locomotive go to the Moon. However, this did not stop railways putting paid to canal transport.
Any business activity which involves analysing data then presenting it linguistically for a particular purpose will be affected by programmes like ChatGPT. This kind of technology is already beginning to do the jobs of humans and will soon do many more.
Let’s take just a few examples.
For starters, there are massive implications for journalism, or at least a certain type of journalism. Throw a press release at ChatGPT and it can rework it into a news story in seconds. The commercial pressures on newspapers means this is only what many human reporters do anyway, especially in local press. Therefore, it is unsurprising that media companies are already considering handing at least some work over to a new breed of ‘robojournos’.
The writers of online ‘listicles’ and blog content also need to take notice. ChatGPT can easily write a blog post about ‘20 ways to make my sourdough less dense’, ‘10 must see attractions in the Lake District this summer’ or ‘15 unusual tips for removing belly fat’. It can scan millions of articles for salient points in the time it takes you to open your laptop. It can write all these articles before you’ve finished your first paragraph. Oh, and it does it for free.
However, one of the biggest impacts could be seen in customer service. We are already familiar with the chatbots who guard the gates to genuine human interaction via the live chat ‘help’ section of many corporate websites.
Imagine these chatbots are able to ‘speak’ like a human and fed on a diet of data consisting of millions of conversations with customers. How long will it be before you are happily text chatting with a very helpful bot who aces the Turing test with flying colours?
How could it affect society?
So, what will happen if millions of people begin losing their jobs to machines?
A dystopian prediction may well say that ‘the Devil makes work for idle hands’ and foresee huge problems controlling a directionless and penniless population, who also have limitless time and energy to focus on innovative ways to cause mischief.
I prefer to take a more positive outlook, an outlook based on our ability to do something machines will never be able to replicate; empathise with other human beings.
In a business context I can readily see a future where the increasing use of artificial intelligence is accompanied by a growing enthusiasm for human contact. Yes, AI may do many jobs and do them well, but the use of human beings in certain roles may become a unique selling point in itself.
Machine learning and AI has the potential to have a massive impact on my own company, Alertacall, where we use technology to provide daily contact to tens of thousands of older people. However, I don’t believe machines can ever replace the friendly voice and listening ear of a genuine flesh and blood human.
We should never forget that our own brains are more powerful than the most complex supercomputer ever designed. We are the ultimate pattern recognition machines, even if we may be unconscious of it much of the time. We are alive to changes in tone, patterns of speech, vocabulary and the context of conversations in a way that machines may never achieve. What’s more, only humans can – for now at least – truly care about other humans and share their experiences in a meaningful way.
Beyond business, could it be that the increasing use of AI spells not a nightmare of aimlessness for society but rather a total reimagining of how we spend our time? Despite its many difficulties, lockdown showed us what humans can do given time and freedom; care for each other, enjoy nature, take things more slowly and embrace simplicity.
Perhaps AI will not render us redundant but grant us the freedom to focus on tasks such as caring for our super-ageing population, connecting with our communities and doing work which benefits the planet.
Could it be a chance to begin measuring our worth not in terms of the money we make, but the good that we do?
Remember you’re a human
Things have come a long way since the young James Batchelor sat down with ELIZA.
And, barring any huge events such as a meteorite strike or solar flare which disrupts our digital existence, technology like ChatGPT will only continue to progress. It seems inevitable that one day, sooner or later, a sentient computer will come into existence.
As a business owner or general citizen of the world, you need to be able to look ahead to these developments and ensure they are something which you are ready to adapt to, work with and perhaps profit by.
But we must never forget that our human ability for critical thinking, empathy, original thought, imagination and creativity will remain just as strong no matter how advanced machines become. In fact, human consciousness only seems all the more remarkable in contrast to the most advanced computers, which are still light years behind.
This gives me confidence that we can develop ways to work alongside machines to improve the lot of humankind and make our world a better place. However, this will only happen if we continue to interact with ourselves and others, to listen, to care and to remember that we have the freedom to change how we think and live.
So, go and give ChatGPT a try, begin thinking about how it can help you and others around you, but don’t forget to socialise, argue with people you disagree with, try and see things from their point of view. Spend time caring about real people and listening to their views and concerns, no matter how wacky or wrong they may seem. Because, in a futuristic world of sentient machines, our ancient human intelligence will be more valuable than ever.