Artificial Intelligence to Generate Business Ideas?

Artificial intelligence (AI) is revolutionising the way we interact with technology, offering creative solutions and even making it possible to generate business ideas in ways that seemed impossible until very recently. To understand this concept, it is important to first talk about the “hallucinations” in generative AI systems.

Hallucinations are a feature, not a bug, of AI systems

What happens when AI “hallucinates” or makes up answers? Hallucinations, far from being a bug, are an inherent feature of generative models, and are fundamental to understanding the creative potential of AI.

The functioning of generative AI models is based on predicting and combining elements randomly to produce content and images, simulating the human creative process, and opening up a range of possibilities to create new business ideas.

The reality is that ChatGPT, for example, is not designed to discern the truth, but to generate content and provide answers to questions in natural language. Everything this system says is based on patterns learned from billions of web pages. Its goal is not to give precise answers, but to facilitate smooth and natural interaction.

Knowing how to speak is not just knowing about any subject

Although this is a topic for another article, it is important to understand that the fact that a machine has learned to speak is in itself an amazing achievement, because these models have not been trained using rules of grammar or syntax of any kind.

But the ability to use language is different from knowledge of a specific field. ChatGPT knows how to speak, but to make it reliably answer technical questions it is necessary to train the system with reliable databases.

The accuracy of the answers will be as good as your database. That is, to give correct answers to objective questions, the model must be trained on specific topics.

AI will never answer anything for which it has not been trained, but what if we set the conditions to usefully exploit the lack of training?

AI to Generate Business Ideas

In the same way that we can leverage training to have accurate answers, under certain conditions we can leverage “untraining” to have creative answers.

Creativity is based on the ability to imagine something that does not yet exist. AI can shine by simulating this creative process to invent new businesses, or new products.

The vision of a product that does not yet exist is a hallucination, and with it all forms of art and creativity that exist. You could say that everything around us has been a hallucination that someone has had at some point.

Could, for example, AI generate the next billion-dollar business? This is the subject of this article on how to use AI to create new business ideas.

In our case we have also set the machine to invent: in HAL149 a model trained with startup business plans invents new projects every day, taking advantage of its ‘hallucination ability’. This system mixes random seed concepts to invent business ideas including the name and skills of the founders.

In this way we go beyond being a content generation tool. We can create a tool to generate business ideas with AI, and the strategies to make them happen.

AI for accuracy vs. AI for creativity

AI giving wrong answers or hallucinating is neither good nor bad. It is in the use and context of the answers that the confusion arises.

• On one hand, only models trained with factual data can give correct answers to technical questions.
• On the other hand, by introducing sources of randomness, we can generate new concepts.

This idea, combined with simulation and management capabilities, opens up unlimited possibilities for creating business ideas and  potentially revolutionary new products.

Creating a new product and building a business around it would be possible if we couple all this to the concept of the AI Augmented Enterprise.

These businesses would be conceived and managed as an integrated and recursive process: capable of simulating scenarios creatively, but also capable of deploying and testing them.

And all this with human involvement that would become merely consultative.