Output Costs Have Approached Zero. So, What Happens Next?

ChatGPT and Gemini have each surpassed 1 billion monthly active users. Gemini generates 150 million images daily, which amounts to approximately 55 billion images annually. Given that Getty Images has a total stock of about 500 million images, AI is producing more than the ‘world’s largest image library’ in just over 100 days.

The same applies to text and music. In the past, outsourcing a single blog post would cost between 30,000 to 50,000 yen. Now, AI can write it in 30 seconds, and a human can edit it in 10 minutes. The effective cost is now nearly zero.

This raises a question: When costs approach zero, what increases in value?

A hint to the answer lies in Spotify’s latest moves.

The Implications of Spotify Excluding ‘AI Artists’ from Recommendations

Spotify has clearly stated its policy to exclude AI-generated music from algorithmic recommendations. While it allows the existence of AI-generated music on the platform, it will be removed from major playlists like ‘Discover Weekly’ and ‘Release Radar.’

Why? The reason is simple: Recommending music with zero output costs would undermine the platform’s trust.

Consider Spotify’s business model. For a subscription fee of 980 yen per month, users are not paying for ‘access to music.’ They are paying for the ability to ‘find music that suits them’ from over 100 million tracks. In other words, they are paying for curation.

If low-cost AI-generated music floods the platform, the accuracy of recommendations will decline. If accuracy declines, users will leave. For Spotify, the exclusion of AI artists is not an ethical issue but a defense of its business model.

This structure is not limited to the music industry.

When Output Becomes Free, the Value of Three Things Increases

As AI’s output costs approach zero, the prices of certain things actually rise. To summarize, there are three:

1. Filtering (What to Exclude)

The cost of creating information and content has become zero. However, the cost of determining ‘what not to show’ has not decreased; in fact, it has increased.

The issue of increasing AI-generated SEO spam in Google search results is emblematic. If search quality declines, users will turn away from Google. Therefore, Google is repeatedly updating its algorithms to eliminate AI-generated spam starting in 2024.

Spotify is the same. Google is the same. We have entered an era where the fate of platforms hinges on the decision of ‘what to exclude.’

Let’s consider small and medium-sized enterprises (SMEs). For instance, if a local construction company produces 50 blog posts using AI, the cost is nearly zero. But how many of those posts will genuinely earn customer trust? The difference lies in whether they can decide ‘what to exclude’ rather than just ‘what to publish.’

2. Context (Why It Was Chosen)

AI excels at creating ‘plausible content.’ However, it lacks the context of ‘why this material was used’ or ‘why this phrasing was chosen.’

Imagine a local sake brewery has AI write a product description: ‘Fruity and easy to drink Junmai Ginjo’ — this is not incorrect. However, if the fourth-generation owner of the brewery says, ‘Our well water never exceeds 12 degrees even in summer. This flavor cannot be achieved without this water temperature,’ that provides context. AI cannot write this.

As output becomes cheaper, the rarity of the story behind ‘why it was created’ increases.

This is good news for SMEs. The context created by a large company’s marketing department inevitably includes ‘corporate logic.’ In contrast, the context articulated in the owner’s own words is raw and irreplaceable.

3. Accumulation of Trust (Who Is Saying It)

As AI-generated content floods the market, the weight of ‘who is communicating’ increases. Even if there are 10,000 anonymous AI articles, they cannot compete with one article from a visible expert with a name and face.

Consider a local tax accountant’s office. Using AI, they could create an ‘explanation article on the invoice system’ in five minutes. There are already tens of thousands of such articles online. However, if a tax accountant who has been in the area for 20 years writes, ‘We actually had this kind of mistake with our clients,’ that becomes unique content.

Trust cannot be built overnight. This is why its value increases in an era of zero output costs.

SMEs Should Focus on ‘Cutting’ Rather Than ‘Creating’ with AI

To summarize the discussion so far:

• AI’s output costs have approached zero.
• When output is free, the values of filtering, context, and trust increase.
• These three are more accessible to SMEs than to large corporations.

Larger companies, due to their size, take longer to make filtering decisions. Their context becomes diluted, and trust relies on ‘corporate brands,’ making individual faces harder to see.

In contrast, SMEs can make quick decisions. The owner can immediately decide ‘what not to publish.’ They can convey the context from the field directly. Trust in the community is tied to their name and face.

Thus, the proposal is this:

Do not use AI solely as a ‘tool for creation.’ Use it as a ‘tool for cutting.’

Specifically, this means:

• Have AI draft 10 blog posts. Select only those that incorporate ‘your company’s context’ and have a human finalize them. Publish two.
• Have AI create five variations of product descriptions. Discard those that ‘do not fit your style.’ Add real-life anecdotes to the remaining one.
• Have AI conduct a competitive analysis. Remove any data that is ‘irrelevant to your customers.’ Make decisions based only on the remaining information.

It’s about selection, not mass production. It’s about editing, not generating. The value lies in ‘filtering through your company’s lens’ rather than ‘using AI output as is.’

Companies That Have ‘What AI Cannot Create’ Will Win

With ChatGPT at 1 billion users and Gemini at 1 billion users, in a world where a total of 2 billion people use AI, there will be no differentiation in the output itself. Using the same tools will yield similar results.

So, what will create differentiation? The ‘unique information about your company’ that you input into AI and the judgment of ‘what to retain from AI’s output.’

Spotify protected the value of its platform by deciding not to recommend music created by AI. SMEs can do the same. The decision not to publish what was created by AI protects trust.

In an era of zero output costs, the competition has shifted from ‘what to create’ to ‘what not to publish.’

Does your company possess something that AI cannot create? If so, that is your greatest asset.