My predictions for 2025 about AI

My predictions for 2025 about AI

The year 2025 just started not even a month ago, and I thought that I might try to guess where the world is going this year. I might be accurate to some degree, I might be completely wrong too, but it's worth outlaying my thoughts on some topics.

AI is a big subject and it will remain big, and it will become even bigger. It will start to crawl its way into multiple areas, and change some industries.

AI Agents is the new hype

This is a term that started to be thrown around a lot, and I think it is going to be the focus of companies in 2025. Many companies will start to develop and sell AI agents that are specialized on various work branches (ex. an AI agent as a personal assistant that will manage your bookings and calendar, an AI agent that will manage some social media accounts on your behalf, AI agent that will research the internet in search of leads, AI agent that will manage company internal documents, etc).

But AI Agents as they are now, in order to become more mainstream will need to develop some new capabilities that they currently lack:

  1. Have access to tons of tooling because they won't be able to interact with external systems reliably. But if they have some tooling that simplifies their interaction with external systems then their work will become more reliable (eg. managing documents on Google Drive will require some way to interact with their API, and will include granular specific actions such as finding a certain file, extracting information from files, downloading files, uploading them, renaming them, moving them around, etc). I don't think the paradigm that AI companies seem to want to push will hold (AI Agents being able to use browsers like a real person and interact with any website/application you put in front of them, that won't happen, at least not this year and not at a level reliable enough to encourage mass adoption).
  2. Do work in the background. Right now, agents and AI systems work on a chat-based interface where you give them some instruction and they execute it. But for AI agents to become a game changer they need to be able to "activate" themselves in the background and do some work without you nudging them with "do this" or "do that". Just think about how we, as humans, interact and instruct each other to do things. We almost never expect some things to happen once just when I ask it, and instead we kind of give instructions and wait for them to happen and get back to us when they are complete.
  3. Multiple communication channels is a must. Right now, most AI apps are basically some glorified chat interfaces. But we, as humans, we talk on all kinds of channels and we expect AI agents that we eventually pay for to be able to communicate in the same way, and eventually among them. Traditional communication channels will need to get integrated in how we interact with AIs (eg. email, WhatsApp, SMS maybe even voice channels such as calling a robot directly to give them instructions). And the communication will need to advance from one way to being two way: you start conversations with the AI, and they come back to you with answers or ask for clarification/help when they are stuck.
  4. Build and execute more complex workflows: Right now, for AI, tooling interaction is limited to basically choosing a tool and using it with the correct parameters. But it's very far from what we need AI systems to do. Usually, when we interact with other human beings we ask them for more complex things, which might be broken out in a full fledged algorithm.
    For example, I would need to instruct an AI agent the following: "When I send you a pdf file, if it is an invoice for My Company LTD then you have to rename it to include the invoice number, upload into the My Company LTD drive into the folder with invoices for the month the invoice was issued for". It's pretty complex and it's a combination of multiple kinds of actions: extracting information from the file, renaming files, accessing Google Drive, identifying the correct folder, uploading there. Maybe it's internal generated algorithm would look something like this
f = download_file(url_attachment)
if check_file(f, "is it an invoice for My Company LTD?"):
  new_name = extract_from_file(f, "invoice number)
  invoice_date = extract_from_file(f, "invoice date")
  rename_file(f, new_name)
  drive_folder = identify_drive_folder("invoices for My Company LTD for " + invoice_date)
  upload_file(drive_folder, f)

   

possible generated algorithm of tool calls to accomplish what I asked it to do

So for 2025, the hype will be AI Agents, and I think there will also be a very sudden rise in companies selling the "shovels" of this hyped trend.

Software that will offer quick access to all kinds of tooling will see a surge of demand. Same for companies that will facilitate two-way communication (monitoring and offering two-way communication through multiple "classic" channels is somewhat of a challenge.

The music industry will change

With the rise of AI music that will become harder and harder to tell apart from real music, the distribution channels for music will see a shift. Spotify already started to recommend AI music in their algorithmic feeds (Weekly recommendations or Song Radio's) and Suno.ai released a new model that sounds cleaner than the previous version and in a few more iterations I think the quality will become almost indistinguishable from real music. For instrumental only music, I think we are already there.

Here is a sample of what it can do:

As this happens more and more, I think

  • Streaming will become less and less a viable distribution channel for bands and musicians. It will offer some visibility, but with music becoming very easy to produce thanks to AI, it will be overcrowded. Music will become a commodity.
  • Musicians' and artists' monetization will become more and more steered towards live performances, which will see an increasing demand. People will crave more and more real life experiences due to AI and internet fatigue (I'll touch this subject more next)
  • Due to needing to be more and more in front of people, artists will need to incorporate more often theatrical experiences to differentiate between them, and their members online personal brands will become the main "selling" channels for their music.

In gaming, it will be the year of the indies

Due to the excessive corporatization of the game industry, more and more AA and AAA titles flop (we see so many examples nowadays), because gamers are fatigues of basically the same games getting re-re-re-released with a high price tag and some small changes to the actual game. And they grow increasingly fatigues of microstransactions and season passes, and will want to get back to the original gaming pricing strategy of paying once and having access to the full game.

Due to all the recent flops and the contraction of the tech industry, many employees from the gaming industry got laid off. This in combination with the fact that AI is making game assets more and more accessible at a decent enough quality, we will see a surge of indie games coming into the market.

This will also lead to a shift in the gaming market sentiment that will steer away from AAA towards indies.

In-person events will see a surge

During the pandemic, we pushed ourselves deep into the online. But now with the rise of AI and everything that happens around the world (especially with the degradation of social media platforms), we grow increasingly fatigued of online and start to look more for in-person events.

We will start to see a surge of real life events, smaller and bigger, and I think this feeds into the section about music: more artists will become performers as well, and a lot of demand for these events will be created. I think we already see some of this with the recent big concerts that get sold off almost immediately, even at insanely high price rates.


What do you think? Am I going to be close with some of these predictions, or am I completely misunderstanding the direction of the tech landscape right now?