One possible outcome of AI's future is rarely discussed to win cash in the gold rush of generative artificial intelligence. If technology doesn't work well enough to replace colleagues, will companies fail to use AI wells or most AI startups fail?
Current estimates suggest that large companies are facing a revenue shortage of $800 billion.
So far, Genai has seen minimal productivity gains, mostly for programmers and copywriters. Genai does something neat and useful, but it is not yet the engine of a new economy.
It's not a bad future, but it's not the one that's currently driving news headlines. And it's a future that doesn't suit the story that AI companies want to tell. Hype burns new rounds of investments that promise large future profits.
Maybe genai turns out to be unworthy, and maybe it's okay.
Is it essential or can't you defend?
Free Genai services, as well as cheap subscription services like ChatGpt and Gemini, cost a lot to run. However, at this point, there are increasing numbers of questions about how AI companies make money.
Openai CEO Sam Altman is openly about how much money his company will spend, and every time ChatGpt says “please” or “thank you,” it costs the company millions. It's anyone's guess how many Openais are losing per chat, but Altman says he'll lose money even with paid Pro accounts due to the high computing costs associated with each query.
(AP photo/Le Jin Man)
Like many startups, Genai companies follow the classic playbook. Burn money and lock in with killer products that you can't afford to miss. However, while most tech giants have not succeeded in creating high-cost products, they have succeeded by preventing them from quitting low-cost products, which are primarily funded by advertising.
As companies try to find new value, the results are “Enshittification,” or gradually diminishing the platform over time by journalist and author Cory Doctorow. In this case, enshittification means that the number of ads will increase to compensate for the loss of free service provision.
Read more: The Internet is worse than before. How did I get here and can I get back?
Openai is considering bringing advertisements to ChatGpt, but the company has said it is “very thoughtful and classy” about how this will be done.
It's too early to determine if this playbook works with genai. Advertising may not generate enough revenue to justify the massive spending needed to make it work. That's because genai has become a kind of responsibility.
Hidden costs of AI models
Another looming issue with genai is copyright. Most AI companies are sued for using content without permission or enter a costly agreement to license the content.
Genai “learned” in many suspicious ways, including reading copyrighted books and scraping off almost everything that is said online. One model can recall 42% “from memory” in the first Harry Potter novel.
Read more: Canadian news media is suing Openai for copyright infringement, will they win?
Companies are exempt from copyright disasters and face major financial headaches of lobbying to repay publishers and creators to protect their models.
American AI startup pseudo-proliferation attempted to pay the authors to train the model to approximately $3,000 per book, summing up to the proposed settlement, which added up to US$1.5 billion. However, it was quickly abandoned by the court because it was too simplistic. The current US$183 billion valuation of herbicides could be eaten fairly quickly in the lawsuit.
The end result of all this is that it is too expensive for AI to own and is like a toxic asset.
Cheap or free genai
Perhaps strategically, Meta has released the Genai model Llama as open source. This means that anyone with a decent computer can run their own local version of Llama for free, whether they were meant to confuse their competitors or inform them of a different ethical stance.
The open AI model is another corporate strategy that locks market share with curiosity bedded side effects. They are not as advanced as Gemini or Chatgupt, but they are sufficient and free (or at least cheaper than the commercial model).
(AP Photo/Jeff Chiu)
The open model confuses the high ratings placed in AI companies. Chinese company Deepseek has instantly tanked AI stocks, releasing an open model that was executed similarly to the commercial model. Deepseek's motivation is vague, but success contributes to growing doubt as to whether Genai is as valuable as it is supposed to be.
Read more: Why building a big AIS costs billions of dollars and how Chinese startup deepseek dramatically changed calculus
The open models – these by-products of industrial competition – are ubiquitous and easy to access. With sufficient success, commercial AI companies may find it difficult to sell services to free alternatives.
Investors may also be more skeptical of commercial AI, which could potentially dry out the seed money tap. Even if open access models are sued for forgetting, it will be much more difficult to remove them from the internet.
Can I own AI?
The idea that genai is unworthy may recognize knowledge. The best genai models are trained from world knowledge. This is so much information that it may be impossible to calculate the true price.
Ironically, these efforts may be from AI companies to capture and commercialize world knowledge. Prices cannot be attached. These systems may benefit greatly from collective intellectual labour so that they cannot truly own their products.
If genai cannot generate sustainable profits, the results can be mixed. Creators who pursue business with AI companies may have bad luck. If the model is liability, there are no major checks from Openai, Anthropic, or Google.
Genai's progress could also stall, leaving consumers with “sufficient” tools at their disposal. In that scenario, AI companies are less important and technology is a little less powerful. That may be completely fine. Users will benefit from accessible functional tools, sparing them from the exaggerated pitch of another round destined to fail.
The threat that AI is worth less than expected may be the best defense against the growth potential of big technology today. If the business case of generating AI has proven unsustainable, what is the best place for such an empire to fall apart than a balance sheet?
