Hot take: ’’Now AI is the message‘‘

Product manager, who worked at Miro and Revolut speaks out loud about the whole "AI-fication".

AI is everywhere. The Big Tech companies have already started to compete in AI products they've been developing. Do you ever get the feeling that you're doing your job like a robot? And that the robot could do better while you do something more useful and creative?

In a column for Durov's Code, industry-recognized product manager Nikita Pakutin discusses what's happening with artificial intelligence, what to expect next, and why it's needed, after all.

What's happening with AI?

New language and text-to-image models provide a refreshing break from the constant coverage of layoffs and minor updates. The meme about robots unable to draw pictures or write music is now obsolete. Basic creative skills have become a commodity, and we are just beginning to explore the vast possibilities that the new models have opened up.

The media coverage tends to focus on the disruptive potential of new models and their immediate effects. However, established players like Microsoft, Notion, Slack, and Miro are rapidly adopting ML-driven features, rendering many startups obsolete mere weeks after their launch.

In spite of slower innovation, tech giants will likely be fine. Just as media companies relied on cable TV revenue while shifting to streaming, tech corporations can leverage their strengths. Google has data, pre-installs, and user habits; Amazon, Apple, and NVIDIA have years of hardware investments.

Current players might even strengthen their positions. Siri might finally meet the vision of a personal secretary in the pocket, and Meta might find more compelling cases for business messaging.

How mediums change us

In 1964, Canadian philosopher Marshall McLuhan introduced the concept of "the medium is the message," which suggests that the comprehension of an idea depends on the delivery channel.

Whether the light is being used for brain surgery or night baseball is a matter of indifference. It could be argued that these activities are in some way the “content” of the electric light, since they could not exist without the electric light. This fact merely underlines the point that “the medium is the message” because it is the medium that shapes and controls the scale and form of human association and action.

One example is the 1960 presidential debates between Nixon and Kennedy. Kennedy won the televised debates but lost the radio ones. Possible explanation is that Kennedy appeared more confident in a visual medium, while Nixon's superior debating skills worked better for the audio.

The delivery channel can even define the message. HBO, a paid cable channel delivered through the satellite network, differentiated itself with more mature and provocative storytelling, while traditional broadcasters targeted females and families, appealing to advertisers.

Years later, Netflix disrupted the industry again. Infinite digital shelf space allowed for more niche content, while also collecting more data than cable providers ever shared with their partners. TV channels clung to their main revenue driver for too long to stop the new entrant.

In 1988, McLuhan came up with ‘McLuhan’s Tetrad’, suggesting that we should ask four questions about the new medium or tool:

1. What does the technology enhance or amplify?
2. What does the technology make obsolete or replace?
3. What does the technology retrieve or bring back that was previously lost?
4. What does the technology reverse or flip into when pushed to its limits?

For instance, social networks have enhanced the speed of spreading information while reducing the role of traditional gatekeepers, such as newsrooms. They have also retrieved the sense of community by recommending content that reinforces pre-existing beliefs. As a result, more people seek validation from like-minded peers, and fakes are mixed with facts, even if quickly disproved.

Why AI will change us, too

It is interesting to consider how society will change when generative content becomes more widespread. While many people have been worried about machines replacing drivers and couriers, it seems that knowledge workers should have been more concerned.

We will find new ways to consume existing content. Text is great for skimming and indexing, while video and audio require more time and focus, even at double speed or with subtitles. What if you could quickly ask for the speaker’s take on any subject?

Text-to-speech, voice cloning, and digital avatars make this possible. Lenny Ratchitsky has built a chatbot from his newsletter archive, and there are experiments with celebrity-based chatbots. It’s likely that every educational platform will soon follow suit.

Generative content can also transform how we view narratives. For instance, we could ask follow-up questions for complicated works like the Bible. Another possibility is to ask Netflix to summarise a previous season of a TV show or even spoil the ending if we lose interest.

‘Professional content’ would be redefined. Media outlets, spammers, and bots could automate their operations with equal success. Startups are already automating writing replies for automatically generated Twitter threads and LinkedIn posts.

A positive example is asset scaling. Red Dead Redemption 2 took about eight years, involved 2000 people, and cost $400 million. The cost of generating countless dialogue lines, locations, and objects will be reduced, and indie developers could create more expansive worlds.

Another opportunity is the digital layer on top of the real world: Google Maps already allows one to look inside buildings with Immersive View, and Apple Maps supports smooth AR street navigation. This could be taken further, simulating a city from decades ago or previewing future developments.

Lower production costs mean that high-quality assets will no longer be a differentiator. In the past, the cost dyes made red and purple colors "royal," and the plastic were initially designer objects. New tools are leveling the playing field for both newbies and established studios, as we have seen in the past with TV and blogging.

We will rethink what ‘trustworthy’ means. Over the last few decades, we have witnessed a decline in trust in authority across commerce, politics, and education. This process is likely to speed up.

Bing's chatbot has proven to be scarily good at gaslighting and manipulation, and it is easy to imagine social media filled with convincingly-looking bot debates. Another possibility is a shadow arms race in creating unrestricted AI engines, both private and state-owned.

Search engines would have a harder time separating quality sources from ML-written, SEO-optimised, and error-riddled pages. This could benefit established brands like the New York Times, as language model creators may prioritize reliable sources and rely on external fact-checking.

In some instances, giving the formal 'he said, she said’ journalistic treatment to answers would not be enough. Tech companies may either need to take a stance on controversial issues, or delegate authority, as Meta with Oversight Board, or Twitter with crowdsourced fact-checking.

Knowledge snippets in Google Search already struggle with facts, and voice assistants struggle even more. Perhaps, knowing that computers can hallucinate would counterintuitively trigger us to fact-check everything, rather than put faith in ready-made answers.

The personal angle would matter more. Content marketing is dead; long live content marketing. Consumers are looking for authenticity and connection with the humans behind the products they use.

Startups like The Browser Company and Nothing Technology turn their CEOs, developers, and designers into public speakers, being more honest and vulnerable than large corporations like Apple who communicate with the world through polished press releases.

The online reputation would matter even more, and curation could increase its importance as a discovery channel. Tiktok and Reddit are already replacing Google for some search requests, and there’s value in the expertise of the outlets like The Wirecutter or YouTube tech reviewers.

More people would become solopreneurs. Webflow, Shopify, and Airtable made it easier to test business ideas. Design, content management, sales, support, ad testing, and coding required teams. Today, anyone can afford the equivalent of countless junior specialists in these domains.

The overall quality of life might improve. Automated tutoring, training, basic counseling, or even legal or medical advice will likely become a reality. While the automated version may not be as good as a human, it is better than nothing. We've accepted the risk of auto-translating the articles and searching for symptoms online: the cost of a mistake will define the value of accountability.

The implications on memory could be wide-ranging. While Apple's Spotlight does a decent job indexing file names and contents, it cannot answer complex queries. Startups such as Findo attempted to solve this but may have been too early.

Startups like Rewind index a history of every conversation, message, movie, and article which had appeared on the screen, both in business and personal contexts. This could augment human abilities and enable us to remember everything or process more experiences at once. Of course, Black Mirror covered the dangers of this 11 years ago.

In the best-case scenario for the future, generative content will free up resources for creative problem-solving, just as typists and telephone operators found new occupations after the rise of personal computing. ML has the potential to transform knowledge work in the same way that the industrial revolution changed manual labor.

As new companies emerge, alarmist takes are shared, and regulation attempts are made, the future is likely to be intense and chaotic. Amidst this, empathy and personality will become even more critical. Perhaps, as a result, humans will have more time to focus on being human.