OpenAI is concerned about the gap between AI capabilities and applications

In a report, OpenAI warns that the gap between what AI can do and how AI is currently used must be reduced. Otherwise, there is a risk of nothing less than a loss of prosperity. AI can perform tasks faster than humans and make people more efficient, the report states.

So far, so AI PR talk. And certainly often true. However, OpenAI apparently does not want to acknowledge that artificial intelligence may not yet be ready for deployment everywhere. And the increased use would primarily benefit OpenAI itself, which is also not addressed in the report.

The report (PDF) states: “However, improved capabilities alone do not lead to productivity benefits or economic impact.” For this to happen, AI tools must be integrated much more deeply into workflows by companies and countries. OpenAI supports the claim that increased usage has a positive impact on productivity with another report it released a few weeks ago.

Once again, OpenAI equates the use of thinking capabilities in ChatGPT with someone being more productive than individuals who ask the chatbot only simpler questions. “Leading countries use three times more deep reasoning capabilities than countries that use it less,” OpenAI writes. Reasoning capability consists of the number of reasoning tokens used in a country.

Of course, it should be considered that countries have access to different AI models whose usage OpenAI cannot see. What can actually be deduced from the report is that, for example, more requests for coding are made to ChatGPT in Singapore than in other countries. The use of AI models can certainly have a positive impact on a company's productivity, perhaps even on a country's economy – but it cannot be derived from the figures cited by OpenAI, let alone a causal relationship.

And the following figures from OpenAI, which are supposed to show that more people need to use more AI functions, also leave questions unanswered. According to them, 19 percent of ChatGPT Enterprise users have never used data analysis. In OpenAI's view, this is problematic. However, these 19 percent may also work in a field where they do not need or cannot analyze any data.

OpenAI wants to become “too big to fail”

The financial difficulties of OpenAI are well known. Subscription models are not enough to cover costs. Now advertising is to be introduced in ChatGPT. Something that has also made other companies like Google and Meta huge and rich, but which users do not necessarily like. In the case of AI and chatbots, users have alternatives – including from the aforementioned companies Meta and Google.

For OpenAI, there is a great opportunity in becoming “too big to fail.” If enough governments and companies rely on its services, they will also do a lot to ensure that OpenAI does not go bankrupt. Switching would be costly and time-consuming, after all. “Our North Star is that AI is treated as indispensable infrastructure and that broad access to AI can be granted to as many people as possible.”

Accordingly, OpenAI also advertises the “OpenAI for Countries” program in the report. The company wants to cooperate with countries and companies and offer them its services, tailored to the needs of each country – entirely unselfishly, “to make the benefits of AI available to even more people.” The program includes support for the introduction of AI in educational institutions, which is now a separate area under the name “Education for Countries.”

In addition, the company wants to offer certifications and jointly develop cyber security measures, as well as be used in healthcare. OpenAI would thus gain access to almost all critical and future-oriented infrastructures of various countries.

In Germany, there is already a partnership between OpenAI and SAP, which also concerns the use of AI for government purposes. “The goal: to give civil servants more time to focus on people, not paperwork, and to ensure that access and benefits are broadly shared.”

(emw)