The Shift: No Accounting For Taste

People tend to judge where something came from before they judge the thing itself. A genuine Monet was posted online mislabelled as AI, and viewers lined up to explain its flaws: incoherent depth of field, muddy colours, reflections that ignored any sense of space. It was Monet all along. This is source anxiety, the instinct to ask whether a machine made something before asking whether it is any good. What the reaction really exposes is how much taste is shaped by culture and by what we assume others will think, rather than by an independent read of the work in front of us.

Taste is judgement that arrives before you can explain it. It is knowing a song is yours within the first two seconds, or pulling something off a clothing rail without quite knowing why. You can choose a suit yourself, put it on, and instantly feel it is not you, with no single reason you can point to. That gut reaction is taste, and it resists being broken into rules, which is why it is so hard to define, to teach, and to reproduce.

AI struggles with the kind of judgement good taste requires, and the reason is built into how it works. Working probabilistically pulls it toward the average and away from the risk and unpredictability that distinctive creative work tends to depend on. The models are measured on coding and the sciences too, so their strengths lean toward usefulness over originality or tone. AI can fold a protein, as DeepMind showed, but the leaps that define culture, like punk breaking in the seventies, come from human judgement that would never have looked like a safe bet at the time. Taste takes risks, and risk is the thing these systems are trained to iron out.

Once speed and efficiency are available to everyone, they stop being a differentiator. When every business can produce work faster and cheaper with the same tools, clients see the hours drop and expect the price to follow. What is left to compete on is the harder stuff to copy: brand, judgement, tone of voice. Most of the current conversation about AI in business is about cutting cost and time, and far less of it is about the quality of what actually comes out the other end.

Taste can be built over time, the way your sense of style at thirteen looks nothing like it does now. That same accumulated experience is what lets you judge whether something is any good. A study in law found AI raised junior lawyers to a solid baseline but no higher, because they lacked the experience to tell whether the output was strong, while senior lawyers could. AI can lift someone to competent, and telling competent apart from genuinely good still rests on judgement that people build up over a career.

Jesse: Hello and welcome to episode four of The Shift, Accounting for taste, where we're going to talk about taste. And more specifically, if I can account for it.

Ben: Yeah. And we have a bit of a theme. The past few years we've talked about the best pizza in New York. We've talked about Jesse's favourite sushi roll. It's a dragon roll. Yes, but that's not the kind of taste we're talking about today.

Jesse: No, no, no. Definitely not. We're talking about whether AI can make static judgements. What is taste? It's a bit more of a loose episode. Taste being such an abstract concept. But also whether organisations can account for taste or invest in taste. Now that speed and efficiency are becoming table stakes, they're becoming a given to stay competitive in the one where the taste is something that businesses need to start thinking seriously about.

Ben: Yeah, I think purposefully investing in and thinking about, you know, when when processes get automated, when things get faster, you know, what is there left to have really? If every organisation reaches the point of optimal efficiency, what's left? What's your point of difference? Anyway? Jump jumping ahead. So we're going to bring it back to a post that we saw this week actually on X.

Jesse: Yeah. This is what spiralled our thinking about the topic of this episode. So and we had a good laugh about it earlier. It was a and I'm actually going to to to to to not spoil it for our listeners. A painting of a monet painting. I did not AI generated who knows. It was posted on Twitter. And the person who put out the tweet invited the viewers to pick apart why. It was it was inferior to an actual, authentic Monet. Here's an AI generated image of a monet. What? Tell me why and give me a good reason as to why this is inferior to a real Monet. And the flood of comments and replies that dissected it, or the depth of field actually have one here. And it went very viral, by the way. Disappointed. I have to disappointed. I have to even point it out. There is no cohesion to the depth, the colour choices. The reflection of the tree bleeds into the lily pads with no regard for spatial depth or for contrast. And that's just one flat, soulless and incoherent model of greens and blues. People were really laying into this AI generated Monet image.

Ben: I mean, you said you're not going to spoil it for the listeners, but I will, because I don't know how we conclude this without doing so. And it was a real Monet. I mean, and this was the funny thing. And I think it just goes to show how perception can be formed before anyone, even, you know, has a chance to come to the conclusion themselves. Right. So everyone thought, because this person said this is an AI generated Monet, they took that face value perception. AI can't do art. Let's pile in on everything that's wrong with it.

Jesse: Yeah, this is source anxiety that people. I mean, I have it, you know, I get a lot of art on my feed and videos even like cinema. And the comments. This is a I even if it's not AI, even if it is I, it's the first question people ask is whether it's I generate it's not whether it's good or what's what it's trying to say. That's like completely secondary. People are in this. Everything is suspect climate at the moment where it's almost like a witch trial where AI content, is being sniffed out. And it's, for me an interesting space because what is this? Go. You know.

Ben: I think and you know, what I find interesting about it is you kind of touched on it at the top of the episode, like, taste is such a abstract concept in a way. And people's reaction to, to things they see, art, video, whatever it may be, is, is driven by taste. Right. And I think what the Monet example proves is that taste is shaped more by culture, people around you and what you think people will think of you for whatever judgement you make, rather than people actually, you know, independently arriving at that judgement of taste themselves.

Jesse: Yeah, taste is so complicated and there are so many definitions floating around. I mean, before we go on to that, people make their judgements of our not just because they know it's made by human, but because naturally, the aura of an art is enhanced by whether you think there was intent behind it. And I think that taste judgement. Is Susan Sontag position is that it's pre rational, right. It's a decision you make about why something is good or bad. Not because it's someone's telling you it's good, but because through ambient experiences and, like you say, cultural influences, you have a sense that something is aesthetically right. And this is why pace is no longer subjective. But for me, it's an interesting area for AI, which is so probabilistic, but is starting to enter the world of serving up creative and serving up, using copy things that require aesthetic judgement are now becoming the realm of AI. So the question really remains, is AI able to make these aesthetic judgements if it doesn't have the experience and the, the, the perspective to shape taste?

Ben: Yeah. I mean, we were talking we were talking about this earlier and I think it's really difficult one to articulate in a way. But the way we went around, it was like you think of any great artist, whether it be musical, performance art, as in, you know, painting or sculpting. You know, most of the people you think of are quite big characters, right? They're not they're not characters that default to the mean or default to the average. Yeah, like AI does. You know what gives them the ability to create something that is so appealing and so different is the probably disconnected way in which their mind works in the first place. And that isn't something that's naturally or AI is naturally optimised to do.

Jesse: It is interesting that AI can fold a protein. You know, Google DeepMind successfully did it, but it can't put together an intriguing poem or a novel that moves people. And I think the traits, as you say, that AI, you know, business owners and companies are controlling for are traits that, you know, make them good at coding or the natural sciences, which are the way that these AI models are assessed. And the benchmarks that they're given in terms of their efficiency are going to tilt towards whether it's useful in a corporate setting or whether it's successful at coding or whether it can help in the lab. Whimsical, even though it creates a unique tone of voice, is also, requires a non predictability that is misguiding in the context of AI.

Ben: Yeah. And I think that, you know, all of the models have purposefully optimised errors, unpredictability. And I guess mistakes or hallucinations out of the models. Right. So so you take you kind of taking that away from it, but knowing that AI isn't going to produce the next Monet is kind of fine like that. That's that's okay, because you can use AI and what you know, it can do. But I think as AI produces more and more within an organisation, having the people in the organisation trained on taste, on what it is that's being produced, that's worth taking forward in whatever context that may be, is going to become an increasing skill.

Jesse: Yeah. And as AI begins to take more and more remit in the pipeline of work, you know, not just in the creative industries but in all industries, this ability to to judge whether something is works, it's something you can't really, as we said before, it evades categorisation, because good taste takes risks. You know, creativity takes risks. Whereas AI is going to avoid risk and it's going to lean towards probabilistic outcomes to think about, you know, be pop in the 40s or punk in the 70s. These are massive shifts in how people consume. But if you trained AI on the date of its day, it would never have bet on the Sex Pistols blowing up, for example. But they did. And it took root because people's aesthetic judgement or their taste was shaped by, you know, their own individual experiences.

Ben: And, and the wider context of the world around it. And, you know, at that time, you know, it's where it's like it's like fashion, isn't it? Tastes and tasting fashion, taste in music, it evolves and it changes. And while today there's this huge kind of adverse reaction to anything that people may have, think, may think has been generated by AI, that's not to say that's going to be the case in three years time. Maybe AI generated content will be actually broadly accepted by society. Maybe it won't. We don't know.

Jesse: Maybe there'll be tacky AI and stylish use of AI.

Ben: Exactly, exactly. And I think it's a really important thing for businesses to start thinking about now. You know, so much of the narrative around AI in terms of business use, anyway, is driving to process efficiency, reduction of cost increases speed. Very little of the conversation is around improvement of output, improvement of customer experience quality, or thinking about the culture within the organisation and the people you need within the organisation to be successful in this new AI era era. You know when skills you know someone that's amazing in Excel. You may not need that anymore because. Because AI is amazing in Excel, it's probabilistic. It's you know, it's very, deterministic in the way it works. So the the makeup of the future employee, when it becomes more about, okay, what to ask, what to do, where to go next, and what's worth taking forward that comes to looking for people with good taste and good judgement, rather than potentially just the good skills.

Jesse: Yeah, it makes me think of I don't know why it makes me think of it, but the Henry Ford quote where he said, if I listen to everyone else's opinion, then I would have tried to get a faster horse. But he obviously built a car and like all his competitors. Oh, we need to build a fast, faster horses to compete with the car. The forward thinking this, and moving away from the status quo is something that businesses are going to need to think about moving forward. These skills that are sort of wrote and trainable efficiencies will will begin smoothing them over. Now, for the most part, I think they want to they want to fear Monga here. But as speed and efficiency becomes a priority for businesses, the, the gains that they see are going to come go in the next bidding quarter because clients are going to begin to see the bill and say, oh, you're taking five less hours to do the same work, and they're just going to they're just going to lower. Yeah. They spend.

**Ben: **Well, well, not even just a competitive landscape, you know, like speed and efficiency will become, you know, a requirement just to stay in business just to compete. And then your differentiating factor again kind of becomes what it's always been. You know, uniqueness, judgement, taste. Yeah.

Jesse: Tone of voice. I dare say something that.

Ben: Yeah. Brand obviously. Because you know brand. Invest in your brand.

Jesse: There's always a salad. Yeah. But I, I think maybe we should move on to what taste actually looks like. What is taste in a professional setting? It's a tricky one to pin down, but really, what taste is, is it's a form of pre rational judgement, right? Like, as Sontag says, it's the beat that hits you. When I was song plays and within two seconds, you know this is your tune. Yeah. It's the intro to Champagne Supernova. It's seeing a sliver of a, of a piece of clothing on a clothing rack and peeking out and taking it home. It's judgement that enters your brain before you've even really begun to think about it.

Ben: I mean, I've got a very practical example of this. I think I was I was telling you. So I'm getting married in a couple of weeks and my wedding suit finally arrived. So it's taken them like two months. Then I, you know, I ordered it, I got measured and everything arrived. I put it on, and despite the fact I'd picked it out, my immediate reaction was, I don't feel that good in this. I don't like it. And I think there was no one specific thing that I could tie that feeling down to. Apart from an immediate gut reaction like a taste of like, this isn't me.

Jesse: Yeah, and you can't trace that reasoning framework that led to your decision that this suit isn't the right suit for you, whether it's, I don't know, or in my school uniform 20 years ago, and I barely even recall that memory, you know, all like anything like the shade of navy maybes. Reminds you of a colour that you said.

Ben: It's pink.

Jesse: So is it pink?

**Ben: **Oh it's not. Yeah, it's the colours. Not the problem though I don't think. Well so I don't know. But it was more that gut reaction, you know, it's just. Yeah. It's not me.

Jesse: Yeah. Yeah. And these decisions about what are a bit of you, of they aren't they aren't decisions. You can, you know, it's black box thinking in itself, which is, you know, you can't enter that space and say, this is why I like this. You just like it. Yeah. And you can't exactly trace it back to us experiences. So how was I going to, you know, untangle that that process. Yeah. Judgement.

Ben: Yeah. And how do you how do you teach it? You know, if you're an organisation like us where a lot of your work product is seen, and you're judged on on that, you know, how do you ensure that across your organisation there is this kind of taste layer before anything goes out, you know, is this good enough? Is does this kind of elevate beyond what an AI model could do, or a client could do directly with an AI model themselves?

Jesse: It's a question of businesses. And as a business owner yourself, how do you assess for taste in the in the process of bringing someone new on board?

Ben: Yeah, I mean, it's really difficult. One, I think when you are hiring, say, creatives or designers, you get a very good sense from their portfolio, right? But you yourself have to have a taste to assess their taste, if that makes sense. So, so my taste in hiring a designer on my preferences, maybe very, very different to someone else within the organisation. So you do have to like instil a little bit of an organisational. What wide kind of what is our what is our taste like? Apple is such a brilliant example of having, you know, taste at the centre of all of their products, you know, and a very strong culture around excellence, a very strong culture around how products and how software, you know, even like interfaces within their products should work. But it's a very difficult thing to pinpoint and distil down.

Jesse: Yeah. Well, I it's it's tricky to pinpoint. I mean, we are in a creative industry. So that's a perspective from which we're going to talk about, but in other professional settings it's like what is what is taste in the finance industry like? And what is taste in, banking. Yeah. Who who knows?

**Ben: **I mean, it's probably things like how what's your tone of voice, you know, how do you interact with clients? What's in good taste and what's in bad taste, you know, how do you present the documentation that you're sending out? You know, if there's not a big differentiator in the underlying service that you're providing, I know, take mortgages as an example. You could go to any mortgage provider and let's pretend they all have the same rates. So you're cutting that point of differentiation out. What then are you left to judge you on your left to judge on brand what other people say about them? Maybe. Yeah. And how how you're dealt with, you know, what is the customer experience like. You know, is it easy? Is it online. Do I have to fill out 15 paper forms or not? And I guess taste in the sense of an organisation like that is in deciding whether that's acceptable or not.

**Jesse: **Yeah, it's a value. It's I think it's the values thing. Like, you know, you could transfer this over, but McDonald's for example, they're all about, you know, colour and and innocence and bringing that out of people. You can tell by looking at a photo whether it would work in a McDonald's ad, you don't know why, or a Coca-Cola ad, for example. It's all about family and coming together and Christmas and that sort of thing. You can tell by looking at a suite of photos which one is most likely to align with McDonald's or Coca-Cola. And it's probably the same for, you know, communications or any any a process, for example, does that align with our value of efficiency? If, you know, transparency is at the core of our business and that's one of our prime business values, it's an aesthetic judgement to say this website doesn't say transparency to me, this this website doesn't say efficiency. To me, that's a taste judgement outside the context of. And when I say a static judgement, it can easily sound like you're just looking at an image or a colour. But it's it's a feeling, isn't it? Which is why it is so tricky to pin down.

**Ben: **Yeah. I mean, it's it's kind of like the age old argument we, we have in, in our industry, you know, creative in general is kind of about taste. And it's like it's very difficult to place a value on it. It's very different. Difficult to necessarily, with absolute certainty, ever say what is right and what is wrong for any given brief that comes to us in a way, you know, there's always a level of subjectivity and there's always a level of taste, either from us or from the client in terms of is that is that right for us or not?

**Jesse: **Yeah. A taste drives you to ask the right questions. Right. But also the unaltered elements of a business are relationships that have trust, narrative synthesis, like we said before, making sure that everything from your headline to your web page to your service are all aligned is something that an AI can really make an efficient judgement on. But as I said before, you know, real world context behind your judgement, through your own personal experience or institutional memory, you know, your experience of working in a company, it's something that AI is going to struggle to account for. And you can see this prioritisation of taste, you know, bleeding over, as we said before into the hiring process, it's not enough to just be able to prompt, but being able to judge whether an output is mediocre or not. There's a massive, you know, experiment that they did in the law industry where they looked at juniors using AI versus seniors using AI, and for juniors, yes, it raised them to a baseline. But they couldn't judge the output because they didn't have the career experience. To really, the taste, I should say to, to see whether or not put was was optimal. Whereas seniors, their experience and their singular, you know, through their experience, were able to, you know, we're able to draw on their taste to, judge whether or not but was useful or not.

Ben: Yeah. I mean, I think that it's it's kind of like quite a commonly spoken about thing in the context of AI is that for someone with genuine depth of knowledge and experience, AI is like a superpower, and it takes you beyond what you could have possibly done in a day before. For someone who is maybe early on in their career doesn't have that like depth of expertise or depth of knowledge, yet AI can only move them to the baseline. It can't move them past it.

Jesse: Yeah. So I guess the question is, can you develop taste over time as well?

**Ben: **I think so. You know, I think that. Certainly you yeah.

Jesse: Look at photos of yourself when you were 14, 13 and.

**Ben: **Yeah, but well, at the time though, that maybe that was the style. That's a thing. Yeah. Yeah. That taste evolves doesn't it. I mean I think that's the most important thing to recognise as well. You know, what is in fashion, what is perceived to be modern design, what is perceived to be a good consumer experience or customer experience changes and it changes really quickly.

Jesse: It's why tastemakers in the cultures are so valued. Yeah, put on a pedestal because they basically shape people's judgement about what works and and what they're going to pay for and what resonates, you know, and not just from like a clothing perspective. You know, if someone puts out a brand campaign with this famous celebrity wearing flannel shirts and suddenly everyone's wearing flannel shirts, it's not just that. It's like also the changing state of the culture and how that affects how certain words or ideas resonate with people.

**Ben: **Yeah, I mean, it's amazing how single moments, single cultural moments can cause such a massive shift in the tastes of millions of people.

Jesse: Yeah, absolutely. Cool, cool. Okay, so on that, we'll wrap up this episode of The Shift. So we've covered taste. Whether you should build that as a core capability in your business, we've covered how taste is fickle but also moves around, you know, with the Monet painting, for example. Interestingly, leaving it open ended around how AI is going to shape individuals tastes as well, you know, whether it's going to make us value, with we're trading on territory of a previous episode now, but effort signals as well. Whether people can actually discern, you know, whether something is AI and or AI.

Ben: Thanks, everyone.