I have had several CEO friends reach out to me and ask me these three questions on artificial intelligence, so I wanted to share my thoughts.

WHAT IS GENERATIVE AI?

Essentially, generative AI is a branch of computer science focused on creating machines and software systems with the goal to simulate human cognitive processes through the use of large language models (LLMs) and diffusion models (for visual and audio data) in order to generate responses with words or images.

The most widely adopted generative AI model is LLM (text-generating), which most people know through ChatGPT (the GPT stands for “Generative Pretrained Transformer”). To delve a bit more into how LLMs work, it is important to start with the question, “what is the model’s objective?” In its simplest form, the systems were designed to recognize patterns and predict outcomes.

Prediction and optimization of the best outcome is one of the biggest challenges of AI, putting ethics, law and bias based on the data sets aside. AI is not new and has struggled with this for some time. For example, in 2016 Microsoft launched its LLM AI, Tay (Thinking About You). The goal was for it to learn by interacting with real humans. While some tried to teach it useful/factually accurate information, other users purposely shared incorrect and inflammatory phrases. Unfortunately, within 16 hours Tay applied a mixture of the learnings and responded to some users with offensive remarks, causing Microsoft to shut it down. Fast forward seven years: several guardrails have been created for AI, and Microsoft moved to incorporate Open AI’s ChatGPT into its search engine, Bing.

It is critical to remember that generative AI systems are not perfect, but I believe the businesses that start to experiment with them will increase performance now and in the future. The key is how you use the systems, and to ensure you are actively seeking to reduce the error rate while still having people analyze the work.

A lot has progressed since 2016, and now there are several mainstream AI systems.

  1. OpenAI developed ChatGPT (text-generating), which is being incorporated into Microsoft services and applications;
  2. Alphabet has its LLM, Bard (I believe is the most basic/conservative – hence it constantly states that it is a “language model” in its responses);
  3. For Image-Generating, there is DALL-E2 (from OpenAI) and Midjourney.

WHAT GENERATIVE AI ISN’T

  1. AI may seem sentient (especially ChatGPT 4) but it is important to realize the algorithm is pulling on response patterns from LLMs that it is trained on. Hence its knowledge was initially limited to the data it has been fed (which was a subset of the internet as of sept. 2021) but this has already shifted in the past few weeks.
  2. AI is amazing but it isn’t human. It does not have the capacity to feel emotions or the ability to have personal memories to draw learnings from.
  3. AI doesn’t know truth from fiction. It’s also important to remember one of its goals is to make you happy (and engage with it) so it is prone to “hallucinate” (or as computational linguists prefer to say, “synthesize ungrounded text”). In other words, it will deliver — very confidently and authoritatively — completely incorrect information so that you get a response that will make you content and keep engaging with it.
  4. AI isn’t fully in our control. LLMs have the ability to pick up new skills which AI researchers call “emergent behaviors” which are mystifying to everyone (even the creators of the models).

WHAT YOU CAN DO WITH GENERATIVE AI TODAY

  1. Analyze competitive information — create a program (AKA code) that will pull competitor information and populate an xls document and then have the AI analyze it.
  2. Idea generation for growing your business — try asking ChatGPT the following: “I am a CEO of X company and would like 20 ideas on how to grow my business in the US this year.” The fascinating part is then when you pick an idea and ask it to explain to you how you can execute it and what it will cost.
  3. Generate a list of influencers for your brand — provide criteria (geo, reach, interest, etc.) and you will quickly have a list of influencers for your next PR event.
  4. Create visual product prototypes — for a future collaboration between your brand and another. Then use the visual output to pitch a potential brand partner to do the real-life collaboration with you
  5. Expedite CX responses — have the first draft of potential customer questions and subsequent customer service responses for an upcoming product launch generated for your team, and it can edit the info. Noting many companies (especially fashion) have gone one step further and already have incorporated the AI technology into their chatbot.
  6. Create buyer personas then pose questions to the personas and ask for corresponding research to back up any statements. This is fascinating to test ideas/learn but also needs to be validated with real-consumers.

I think it is worth noting, no company or person (even the developers who created the code) is an expert in the space of Generative AI. As it is evolving, governments and regulatory bodies are still in the process of creating their rules around the use of generative AI tools. Some EU countries such as Italy have already banned ChatGPT due to privacy concerns. In the U.S., regulations are being explored and will inevitably be introduced in the future, but for now it is up to us to explore this new frontier.

Like any explorer, I am very curious — I would love to learn how you are using generative AI to grow your business?

This article was originally published on womeninretail.com and has been shared with permission.