Per and Britt Karlsson already use artificial intelligence quite extensively in their work, but certainly not to write text, nor to create images, which are two of their core vinous activities. So, how do they use AI? Here, Per takes a look at some of the areas where it can come into play.
“Do you use AI?” Today, this is no longer really a question. We use AI, everyone uses AI. You cannot refuse to use AI. It’s a bit like saying in the 90s, “I won’t use the internet,” or “I have no email.” Maybe that approach worked for a while, but in the end, you just had to give in and start using it, and once you did, it actually turned out to be a great tool – provided you’ve become aware of its limitations and use it with care and intelligence. Indeed, soon you realise that you simply couldn’t do your work or live your life without it. You might even be using AI today much more frequently than you think.
We use artificial intelligence for translations; photo and video; transcription; texts; data mining and research; daily life; proofing, spellchecking and grammar checking; and for troubleshooting tools and technology.
But what AI?
Translations
If you’ve ever needed to translate a text, and used a tool for help to do it, you’ve used AI. There was a time when, for example, Google Translate used linguistic logic, analysing sentences and applying rules to translate. Then they changed it to use an AI engine for the translation. At the very beginning, the translations sometimes came out a bit curious. But very soon they were much, much better than before. That change was in 2016, so you’ve probably used AI for quite a while.
Since we, Britt and I at BKWine, write in both Swedish and English, and sometimes French, we use this all the time. It gives excellent results, but it always has to be checked manually. That’s a good description of most things that come out of AI: good results, not 100% reliable, which always need to be checked manually.
There are different levels of AI use here. The simple use of Google Translate, or others: LanguageTool, ChatGPT Translate, DeepL, Hix, Hemingway, SlickWrite, Reverso, Wordreference. But you can also use one of the chatbots (more on those later) for more specific translation work. For example, “translate in a very formal way”, “make this text very enthusiastic”, etc.
Photo and video
We do not use AI to create images. That is perhaps mainly because we are also photographers, so we always only use our own images. We know some people who use it extensively, but the images are often striking in their ‘AI feel’ and underwhelming in their impact.
But AI can probably play a more important role in that area soon, too. But – for us – perhaps more for video. Some AI-based video editing and creation tools are becoming very interesting, and we have done some trials. Descript is a video editor that looks very interesting, though we have not used it yet. Blackmagic DaVinci Resolve is also coming out with its own AI tools soon. And so are others.
What I do use it for when it comes to photography is cataloguing. With a few hundred thousand photos in the archive, you need a way to find them. I use a catalogue program, a ‘digital asset management’ (DAM) tool. The one I use most is called Daminion, though I also use Photosupreme, which is a bit less elegant and less user-friendly to me. No, Lightroom is not really a grown-up DAM. It has a clever feature that uses AI to automatically ‘tag’ or ‘keyword’ the images. This means that they become searchable. I can, for example, search for ‘amphora’ and it will find pictures with amphorae. I can even search for, say, ‘Languedoc’, and it will find wine bottles with that word on the label. It’s not perfect, but it’s tremendously helpful.
Transcription
Another area where artificial intelligence is tremendously helpful is for transcription. This works, for example, for interviews you record or for a presentation you’re listening to, but it also works for videos. Most of these services also manage to handle different languages. The service we use for that is Otter.ai. It’s not free (at least not if you use it frequently), but it is very useful.
Text and writing with AI
This is, of course, the main area of debate and the most contentious for journalists. We work with one media that, in their guidelines, says they “DO NOT PUBLISH TEXTS WRITTEN BY AI.” We agree. Not everyone agrees, it seems.
Sadly, we do see a lot of texts that give the impression of having been written by AI, or with just a very slight human touch-up, and then published under some supposed journalist’s name. We also see a lot of text taken directly or almost directly from marketing materials (such as a wine region’s promotional site) and published as if it were journalistic writing. I don’t know which is worse, the former or the latter. Both are ethically wrong and discredit the wine journalism profession. That should make it clear: we do not use AI-generated text in our articles. But there are several other ways artificial intelligence tools can contribute positively to text creation.
Data mining and research
AI is a powerful tool for background research and, to some extent, for data mining (if you need it). You can ask it any question, and it will come up with an answer – almost always. We use it fairly often, to some extent, as background for what we write on BKWine Magazine and Forbes.com, as well as for travel pieces on BKWineTours.com and for our wine tour activity. (One simple example: finding alternatives to NannyBag in this article: What do you do with your suitcase when you want to go for a walk and can’t find a locker at the station?)
But, and this is a big BUT, it can ‘hallucinate’. You cannot entirely trust it. It can easily get the facts wrong, so any info you get from it should, in some way, be double-checked (and preferably not with another AI chatbot). But for a general overview and quick facts on various subjects, it is very effective. We often use it to get an overview of various subjects or to get inspiration on how to approach or structure a subject.
Some AI tools can also produce quite in-depth research papers, if you ask the right questions. But here too, you cannot entirely trust the facts.
Daily questions
We often ask AI almost any question that comes up in daily life, too. “What happened in 1628?” “How do I best get from Pézenas to Montpellier?” “Is there a better camera app than the generic one on my phone? It’s not always 100% reliable, but often good enough.
Proofing, spellchecking and grammar checking
Artificial intelligence is also very helpful to improve the ‘technical’ quality of your writing. One good tool is Grammarly (www.grammarly.com), which is a spell-checker, grammar checker, style checker, and a ‘general tips for improvement’ tool. Unfortunately, it is only available in English. Sometimes we use one of the AI chatbots for similar things. I’m sure many other tools in this area would be worth exploring.
Technology
One area where artificial intelligence really comes into its own is technology, broadly speaking. If I have a problem with Windows, I ask AI for help. If I am looking for a program to do a specific task, I ask AI. For example: “Are there any good and simple programs to trim videos without having to re-render them?” or “I am fed up with the intrusive McAfee, what other good anti-virus and security programs are there?”
We are also running several websites for our business, built on WordPress and Elementor: BKWine Magazine, BKWine Tours, BKWine Vinresor (our Swedish travel site) and a few others. They require constant maintenance and adjustments. For example, I recently had an issue where the hero image (a video, photo, banner or illustration that greets visitors to a website) on the pages didn’t behave as it should. I asked AI, and it gave me some CSS (Cascading Style Sheets) that fixed the issue.
For anything related to technology, AI is exceptionally helpful. But again, it can ‘hallucinate’.
More?
I am sure I have forgotten some of the ways we currently use artificial intelligence, and that there are some areas we have not yet explored.
But which AI Chatbots?
Some of the tools and apps mentioned previously have their own ‘built-in’ AI, but when it comes to chatbots, what do I choose? I use several different ones. They work differently and give different kinds of answers. And it can be wise not to rely on a single one. The two I use most are Google’s Gemini and ChatGPT. Incidentally, ChatGPT is also the ‘engine’ behind many apps. I find I like those two best.
Both have free and paid versions. We pay for Gemini (as part of Google Workspace) but not for ChatGPT. Gemini has different levels of ‘thinking’: Fast, Thinking, Pro and Plus. Many of the others have similar schemes. Considering the investments behind these systems, it is surprising that so much is free.
Others I have looked at or occasionally use: Manus – is a bit quirky. Claude – sometimes seems a bit outdated. Perplexity – usually gives more references to sources than others. Copilot – Microsoft’s version, which is not too bad, actually, and today built into many computers. Some I have not tried: Elicit, DeepSeek, ChatPDF AI and Humata.
You can never fully trust an AI chatbot – it sometimes ‘hallucinates’ (i.e., produces incorrect, made-up, or ungrounded-in-real-data information). If you use it for something important, you have to double-check. Originally, most chatbots did not provide references to their sources. Now, it seems more and more that most of them do. This is good.
There is no way to avoid artificial intelligence. It is better to embrace it and understand it as much as possible, while remaining aware of its limitations and dangers. Embracing and understanding it also means not abusing it, which is a danger, not least in journalism.
Main image photo by Barn Images