A pessimist sees the difficulty in every opportunity; an optimist sees the opportunity in every difficulty.
You may think AI (Artificial Intelligence) is just for the technical experts. Wrong. Give it a try free at chat.openai.com
First a warning: the information is sourced from a wide range of (mostly online) resources. That can include unreliable sources so before taking it as accurate use your common sense and check any information you intend to rely on with an authoritative source. It should be noted that chatGPT relies on data harvested from the internet in 2021. That will be updated and chatGPT is only one of many AI tools, currently billions of dollars are being invested in openAI and other AI organisations and it remains to be seen which will become the market leaders.
The word “intelligence” in AI is something of an exaggeration, frankly some of the results are not just wrong but dangerously so. That doesn’t mean it’s worthless because much of the output is extremely useful and valuable. If you apply your real human intelligence then the output of AI will very often remind you of aspects you may have overlooked. Think of AI as your office junior; their work needs checking. Alternatively use it for entertainment rather than work as in my examples.
You will need to sign up but there’s no request for payment information, just email and a unique password.
The free trial is limited, if their systems are busy serving paying customers responses will be slower and may quit part way through. The paid version costs USD20 per month. Having tried some experiments I’m tempted.
Try a sentence you might use in Google search and see how well it does. Rather than giving a list of web sites that might help the AI will summarise that information. It works in “chat mode”, if you ask a supplementary question the response is in the context of the first.
Both Google and Bing (Microsoft) search are working on integrating AI into their systems, the difficulty both face is trying to maintain the quality of the information provided. At present most AI systems will present as fact incorrect information it has found somewhere and it doesn’t identify the source.
That doesn’t mean it’s useless, it means you mustn’t place blind trust in the output. If something looks wrong it may well be. Check facts independently.
That’s no different from “real life”, human intelligence is pretty flawed too. A BBC weather forecast on 29 June 2023 showed he same unseasonally low temperature for 4 days in 4 very different UK locations each of which was to experience differing sunshine, cloud cover and rainfall. It was obviously wrong and yet the BBC broadcast without checking.
Is it dangerous?
In common with many other developments it can be used for good or evil. Many technologies we now regard as progress have adverse effects. Let’s face it, were it not for the industrial revolution we’d not now be facing pollution and climate change. AI can be used for evil and as capabilities improve the situation will get worse but individuals, government, the UN are powerless to stop it. What we can do is implement controls and educate people about the risks.
I used this segment as a test of AI. I wrote:
There are very valid concerns about AI but the genie is out of the bottle. We have to learn how to make best use of the resulting benefits while retaining some level of scepticism and being alert to the risks and dangers. AI falls a long way short of human intelligence, fundamental problems include that it lacks understanding, isn’t capable of creativity and isn’t good at cross-checking data to verify information.
It is a very transformative development some have compared with the printing press, steam power, nuclear weapons and the internet. It can be used for good or ill but like those examples, the genie cannot be put back in the bottle. I have used it to generate text documents for a number of purposes. The result was a better document than I would have produced it used good grammar and spelling and included relevant informationI may have overlooked. I did need to do some fact checking and proof reading but overall I’d got a document I could use in a few minutes rather than maybe an hour or so.
As for “use for evil” yes that’s happenning, AI generated spam is already appearing. The online support chat provided by some organisations has been automated for some time but now phone support can do the same, “understanding” your spoken question and “speaking” responses. Those phone calls you get typically purporting to tell you to take urgent action in respect of a financial transaction or delivery are starting to use the same technology because it’s cheaper than using human call operatives and the AI makes them sound more plausible.
Then I submitted that to AI and this is the version it created:
There are valid concerns surrounding AI, but it is now impossible to reverse its development. We must learn to maximize its benefits while maintaining a level of skepticism and awareness of the risks and dangers involved. However, it is important to note that AI falls significantly short of human intelligence. It lacks understanding, creativity, and the ability to cross-check data for information verification.
The impact of AI can be compared to transformative inventions such as the printing press, steam power, nuclear weapons, and the internet. Like these examples, AI cannot be contained once it is unleashed. Personally, I have utilized AI to generate text documents for various purposes. The results have been superior to what I could have produced on my own, with proper grammar, spelling, and relevant information that I may have overlooked. While some fact-checking and proofreading were necessary, I saved a significant amount of time in generating a usable document within minutes instead of an hour.
Unfortunately, the use of AI for malicious purposes is already occurring. AI-generated spam has become prevalent, and even phone support services now employ AI technology to understand spoken questions and provide responses. Scammers are also utilizing AI to make their fraudulent phone calls sound more convincing, leading unsuspecting individuals to take urgent actions regarding financial transactions or deliveries. This approach is cheaper than employing human call operators and increases the plausibility of their scams.
So how did it do? I wasn’t unhappy with the original but the AI version covers all the same content but perhaps easier to understand without the colloquialisms like the genie is out of the bottle and it does switch to the American spelling of skepticism, apart from that I can’t tell that it’s AI generated. I’d be happy to use the result and it only cost me $0.002
Next I tried asking AI to generate the article (rather than revise mine) here’s how that went:
Introducing AI in copywriting can be a double-edged sword. While it brings efficiency and speed, it also poses dangers. AI lacks human understanding and creativity, potentially producing generic and soulless content. Misinterpretations and biases can lead to misinformation or offensive text. Trusting AI blindly risks losing the personal touch and authentic connection with the audience. Balancing AI’s power with human expertise is vital to ensure quality, ethical, and persuasive copy. Let’s harness AI’s potential responsibly, crafting compelling narratives that captivate and resonate, without compromising integrity.
A lot shorter but it covers the main issues.
How does it work?
Chat GPT uses a LLM (large language model). An LLM is a massive database of text taken from various sources, that might include academic research papers, news media, the internet and even works of fiction. This type of AI works like the predictive text you may have on a mobile phone by attempting to predict the next word from the context and previous word. If you type Happy your mobile may present a list of probable next words like birthday, christmas, anniversary. An AI chatbot is essentially a more advanced version of that.
Mastering the Art of Harnessing AI: Safeguarding Your Copywriting Endeavors
I handed the task of writing this to chatGPT with this instruction:
“Write an article for a web site about how to mitigate the dangers of using AI to generate copy in a confident tone and instructive style.”
In the ever-evolving landscape of digital marketing, AI-powered copy generation has emerged as a powerful tool. However, it is crucial to navigate this technology with caution. Fear not, as we unveil the secrets to mitigating the potential dangers of using AI to generate copy.
1. Embrace Human Insight: While AI can streamline the writing process, human input remains invaluable. Inject your expertise, creativity, and unique perspective into the AI-generated content to ensure it aligns with your brand’s voice and resonates with your target audience.
2. Set Clear Guidelines: Establish comprehensive guidelines for the AI system, outlining preferred writing styles, tone, and objectives. This instructive approach will help the AI understand your brand’s identity and produce copy that consistently reflects your vision.
3. Monitor and Refine: Regularly review the AI-generated content and make necessary adjustments. Analyze customer feedback, engagement metrics, and conversion rates to optimize the AI’s performance and ensure it continues to meet your goals.
4. Ethical Considerations: AI-generated copy should always adhere to ethical standards. Avoid plagiarism and ensure that the AI is not generating misleading or harmful content. Stay vigilant and take responsibility for the output generated by the AI system.
5. Stay Updated: The world of AI is continuously evolving. Stay informed about the latest advancements, updates, and best practices in AI copy generation to leverage its full potential while mitigating any associated risks. By adopting a confident and instructive approach, you can harness the power of AI to supercharge your copywriting endeavors while safeguarding against potential pitfalls. Remember, AI is a tool, and it is our responsibility to wield it wisely to achieve remarkable results. Embrace the AI revolution, and let your brand’s voice soar to new heights!
“It’s more dangerous than a viral pandemic, global warming or nuclear war” and “It’s a revolution comparable to the invention of the printing press, steam power, computers, internet, mobile phones”.
Like all those, true or not, it’s unstoppable. Like all those there are two sides to the story, risks and benefits. If we don’t make best use of the benefits all we are left with is the risks. Worse, others will find positive uses and we will be left behind.
Similarly with the negative aspects, if we aren’t alert to the changed environment, others that are will be able to exploit what has become our weakness.
Press and anyone wishing to promote a story for their own benefit will tend to sensationalise and make it a scare story.
The danger is that people will regard AI solely as a danger and may try to disregard it. Ignorance is even more dangerous.
Google’s parent company lost $100M stock market valuation when its AI launch demo went like this:
“What new discoveries from the James Webb Space Telescope can I tell my 9 year old about?” The chatbot response included the claim that the telescope took the very first pictures of “exoplanets”, planets outside of earth’s solar system. James Web launched in 2021 the first exoplanet photo was taken in 2004 from a conventional observatory.
That was a knee-jerk reaction, a misunderstanding of AI a belief that it was going to be infallible. Google’s implementation came as a rapid response to Microsoft Bing Search & Edge browser incorporating an element of AI and the initial demonstrations of that too included egregious errors but they didn’t gain the same level of publicity and the stock price was unaffected.
Examples of AI generated text
I tried asking ChatGPT “Create an itinerary for an activity holiday in Finland“, it gave a page of useful suggestions probably partly drawn from travel companies and tourist information services. That was a better starting point than Google search results but not a replacement. The same search on Google provided a list of potentially useful web sites but it would have taken longer to trawl through those to extract the key suggestions and it gave links to travel companies that offered itineraries that might prove suitable. Next I asked “best time to visit“, that’s a bit of a challenge because it depends what I had in mind with “activities”, winter sports or kyaking for example. The response anticipated that issue and gave a 4-season breakdown of options. A clue to the sources being those promoting travel to Finland was that it said “Summer (June to August): This is the most popular time to visit Finland” but also “Winter (December to March): Winter is the peak season for tourism in Finland” which one might consider to be contradictory but if the data-sources include travel agents or the national tourist bureau they will want to promote year-round travel.
Next I tried something very different “write it in the style of william shakespeare” (because… why not!). I’ll give the response in full because it performed the task better than I expected and it gives a different idea of the capabilities of the system.
Oh hark, dear reader! Pray tell, lend thine ears,
Forsooth, I come to thee with quill in hand,
Inscribed with ink, as black as raven’s feathers,
To pen a tale of grandeur from a distant land.
Perchance, I shall weave a verse, as fair as May,
And paint a picture of beauty, with words so gay.
The world is but a stage, and we its players,
And thus, the story of our lives unfolds,
Filled with joy, and pain, and many a betrayal.
But hark, what’s this? A story I hear,
From a land far away, beyond our reach,
A tale of knights and kings, and their fates,
In a land where winter holds its icy grip.
Finland, I hear thee, a land of snow and frost,
Of northern lights, and wild reindeer’s hooves.
Where the mighty bear roams, and wolves howl,
And the forests stand, deep and dark, as death.
In such a land, adventure awaits,
With skiing and snowmobiling, and huskies to lead,
And fishing in the frozen lakes,
And the magical northern lights, to behold.
Oh, Finland, thy beauty and grandeur,
Doth rival that of any other land.
And though thy winters are long and cold,
Thy warmth and hospitality, we doth hold dear.
Thus, I bid thee, dear reader, farewell,
Until next time, when once again,
I shall weave a tale of wonder and joy,
With words as bright as the morning sun.
The answer is yes! It wrote the program for me.
Then I asked “create an html interface” and it wrote a web page to display the ouput from that program.
To use that generated program in “a production environment” would need more work, but it was a good starting point, saved time on some common routine programming work. Thinking of it as the apprentice or office junior, it’s saved effort by doing the basic work but a supervisor needs to cast a more professional eye over the result.
Being the fool I am, I went further with this test. Having been impressed by the Shakespearian rendering of the travel advice I tried the same on the program code and once again the outcome was surprisingly good and quite amusing (and completely useless!)
Hark! Behold yon HTML interface,
Whose beauty doth shine with CSS grace.
‘Tis made for CRUD, that task so drear,
And yet with PHP and JS, it doth appear.
The form doth sit, like a lady fair,
With labels and inputs, oh so rare!
And a button too, of green so bright,
To add a user, with all its might.
The table doth come, with head so wise,
Containing data, of all sorts and size.
Its columns three, with actions four,
For CRUD is not a task, to do in poor.
And lo! The styling, so elegant and fine,
Like a garment of silk, that doth shine.
The background, a shade of pale grey,
And the box-shadow, like a cloud, at bay.
The font, so royal, of Arial breed,
And the text-align, centred with speed.
The borders, a touch of black so neat,
And the padding, like a cushion, under feet.
The buttons, so noble, with colours rare,
Green and red, like the lion and bear.
And the hover, a touch of opacity,
Like a veil, on a lady of felicity.
Thus doth end, this HTML interface,
Whose beauty, doth outshine all race.
With PHP and JS, it doth converse,
And CRUD, with ease, it doth disperse.
But AI is not just trivia and not just text. The demos above are very basic, limited to regurgitating and reformatting existing information. AI is a massive growth sector and there are already uses in industry, engineering, medicine, art, music, film. Chat GPT is bringing a greater level of awareness to a wider public. Take a look at Google’s collection of AI experiments here experiments.withgoogle.com/collection/ai
The focus here has been on text, what about audio, and images?
You may have heard of this email/text message scam:
You recieve an urgent request for financial help from a friend or family member. Perhaps they are on holiday and they need funds to pay to have their impounded car released, to pay a fine, pay for repairs (there are other scenarios like needing to pay medical fees). Enough information is included to give the appeal some credibility, perhaps it includes their holiday location and you knew they were going away. That information is easily gleaned by the fraudster from Social Network postings or hacked email accounts. A capture of a few seconds speech is now enough for AI to generate a voice message, you could find an answerphone message in the readily identified voice of a close relative, now on holiday asking you to send money urgently.
You may think that’s fanciful – it’s not, it’s coming to your phone any day now, and the technology is only going to get better. It’s a small step from AI text chat to interactive AI voice chat in whatever voice has been sampled.
You need to be alert to the risk to be able to protect yourself. To do that firstly you need to be alert to the risk factors. At present any unsolicited contact involving payment, moving of valuable assets or urgency of action needs special attention. To ameliorate the risk you need to verify some other way, ask a question an AI system is unlikely to be able to answer, even if the fraudster had gone to some lengths to research things like birthdays, pets, make of car. It is worth investing time and effort into that research as scams can deliver £thousands. Two suggestions:
- Firstly be extremely sceptical of any insecure means of sending cash. Western Union was for a long time the fraudsters’ chosen mechanism, examples have been widely reported of payment being made to the recipient with inadequate ID checks. Fraudsters also know the value of the sunk cost fallacy (Google it), maybe asking for a small amount initially then, having established your gullibility, coming back for more.
- Perhaps a bit extreme but perhaps have a family code word pair, let’s say “apples” and “pears” if one participant in a dialogue includes “apples” and the respondent includes “pears” then the dialogue is probably legitimate.
Image manipulation, whether AI assisted or not, has become extremely capable and easy to use. You may have seen the photo of King Charles dressed as a biker and riding a motorcycle. There are tools that can estimate the liklihood that a photo has been AI generated or an original has been modified but the real test is down to your application of your intelligence.
LLM based AI is heavily flawed and it is not comparable with human intelligence. Although some of the output can give that impression and it is good enough to augment some human “creativity” it’s really pretty stupid. But it’s certainly very useful, for example we are now used to spell-check when creating documents, this goes one step further and can help with grammar, style and tone. Good progress is being made with checking for accuracy of content , that won’t be perfect but faster and more thourough than humans.
The next big step is GAI (General Artificial Intelligence) it will be difficult to agree if/when that is achieved, it is intended to be human level intelligence with features like true creativity, insight and innovation and how do you measure that? Arguably that’s when it’s smarter than us!
Just what are the risks?
Some are still imponderable but the nature of study and of many “white-collar” jobs is going to change.
The most imminent risks come from AI generated speech. A short recording of your voice is enough to make you appear to say anything. A few photos of your face enable AI to lip-sync that AI voice and it would be trivial to create a video of “you” or the PM saying something quite outlandish. A greater risk is making that video say something credible, but false, which would be good enough to convince many viewers it was true. Only expert examination could tell true from fake and denials are useless. “A lie can travel halfway around the world before the truth can get its boots on” is the most current version of the quote but alternative versions with the same underlying meaning can be traced back centuries.