Taste and Judgement are lies we tell ourselves

Tech Notes

SubscribeSign in

Taste and Judgement are lies we tell ourselves<br>Dude, token predictors will never have a soul like us. Okay?

Abhishek Anand<br>Jun 20, 2026

Share

I walked into a bookstore and saw someone click a picture of the bookshelf apparently to ask ChatGPT about a suggestion. I have mixed feelings about this.<br>Kafka once wrote:<br>I think we ought to read only the kind of books that wound or stab us. If the book we’re reading doesn’t wake us up with a blow to the head, what are we reading for? … A book must be the axe for the frozen sea within us.

While we are drowning in monocultures of code, writings, ideas, even scientific research, I see writings about taste and judgement everywhere. Blogs, X, poorly organized Linkedin posts sent to bots to fix. Even companies being formed on the premise that slop sucks and taste is what matters.<br>The premise is AI cannot do taste and judgement better than us. Not even in the future. This is what separates us from them.<br>Anthropic annihilating anthropocentrism

Long before I overheard two engineers in the Bay Area discussing how they would ‘solve’ Tenderloin issues with clever incentive design, there were actual polymaths. Descartes and Leibniz drank their tea and coffee and occasionally formed wrong theories on topics like what makes humans sentient.<br>Descartes had this idea that we could reason and talk because we had a soul and animals were mere ‘beast machines’, an automaton in flesh and blood. Leibniz corrected him and agreed animals also had ‘natural souls’.

A Cartesian animal machine image generated using Nano Banana. Look no sentience!<br>Freud’s three blows to human narcissism is discussed a lot recently with AI being added as the fourth. And while reading I found that Freud might have got two of the three ‘inspirations’ from another speech much earlier and then added his own name as the third.<br>To recap, it goes like this. Humans were all happy to be the snowflakes, living in the center of universe with the sun and everything revolving around them. Then this dude called Copernicus comes and proposes the heliocentric model of our solar system where Sun is the center and we revolve around it. To add salt to injury, it was later clarified that we are not even the center of the universe but just some tiny random speck in some corner of it.<br>The people were like okay, but we still are these specially created things with soul and can reason and have emotions unlike the other animals. In comes Darwin. We evolved to become humans, like other creatures. Nothing special.<br>Also, it turns out crows can plan too, elephants grieve and chimps are capable of deception.<br>Then Freud added his own name adding that human consciousness and being in rational control of our minds is a fraud with the id, ego, superego thing.<br>We have not yet agreed on our understanding of what consciousness even is yet a majority of us will deny that a machine or AI can ever be one.<br>When Deep Blue beat Kasparov, one of the greatest Chess players of all time at his peak, the news headlines was ‘Computer beat Kasparov’. Later on, it would reduce to brute force search beat Kasparov, there was nothing AI about it. AlphaGo broke the next one. Go was hyped as more creative and thoughtful than Chess.<br>At the current pace, models will improve to the point that writings, image and video generations do not appear as slop to most humans. We will deny taste and judgement in machines till we can. Because we can and let’s be honest, what does taste even mean objectively? Once the fence is gone and we are overrun, we will retreat and seek shelter in something else. Morality? Soul?<br>It turns out we will not even have the intellectual superiority.<br>But we will always have arts?

Experiencing something like Bi Gan’s Resurrection as I did recently, once in a while does give some weight to the argument. Can a Claude skill be a generational visionary like Tarkovsky? Or even closer to home, a Satyajit Ray?<br>You see a rare calligraphy hundreds of years of old at a museum and wonder if the robot trained with video data from labour exploitation in extremely low wage countries can do the same ever.<br>At the other spectrum we have bad designs in AI. MCP is a disaster. Skills.sh is a graveyard of outdated markdown files without an optimization or update loop.<br>We now live in a world where RentAhuman does not sound a terrible name for a startup anymore. Once imagination as the last frontier is done, nothing remains.

Stalker by Tarkovsky

“Learn to use AI well” does not save your job

Generation has become cheap, verification has not. It apparently still needs humans till the time it is good enough to be automated. Ages ago, I used to look into VS Code for every piece of code the AI agent would generate. With Opus 4.8 and a few days we had Fable, now I just glance over the PR, unless it is something really critical.<br>In the end state, AI will use AI. Most of the people are using...