WEW23 - What is AI anyways?

Another week, another experiment with the format. For the next few issues, I’ll be structuring the content around a single technology, what it is, how it impacts business, and how you can use it today. Let me know your thoughts each week and if you found anything I shared helpful.

The best compliment I can receive is a new subscriber. If you find my weekly emails to be insightful and a platform for your growth, please forward them to at least one colleague you think could also benefit.

If you were forwarded this email, you can subscribe for free!

In this week’s Email:

✧ How AI Works - What Are We Actually Interacting With?
✧ Impact - How Will Businesses Continue To Respond To Generative AI?
✧ You, Today - How You Can Leverage LLMs at Work, Today.
✧ Further Reading - Books and Articles I recommend this week

How AI Works

What Are We Actually Interacting With?

The news cycle and recent advancements make it seem like AI and its child Machine Learning, are brand new. But we have been using them for years, even their generative varieties. Generative AI is any artificial intelligence system that creates some form of content or output. The term intelligence here is misleading, as often these systems are powered by relatively simple math or weights.

When we look at a system like autocomplete for text messaging or email, the simplest form of AI to achieve this is a Markov model. Markov models are weighted trees of data that based on previous examples guess what the next most likely piece of data would be in a sequence. It doesn’t have to be words either, it could be any sequence. Six years ago I leveraged a Markov model to make glitchy Bauhaus art.

The current emerging tools in AI, which are Large Language Models (LLMs) work differently from something as simple as a Markov model tracking weights for which words are most likely to be next. These systems are networks of nodes storing a numeric weight, that process what an output should be for a given input. The exact way they work varies from implementation to implementation. The common ground is that they still pick what words come next in a sequence, but do so with context and accuracy vs pure linguistic probability.

Unlike a Markov Model, which is easy to inspect and understand these networks of weighted nodes (Neural Networks) don’t lend themselves to human interpretation. This means that while we can guess how a system was trained to get to a certain type of output, we aren’t able to inspect its process of output selection. But regardless of how advanced these systems like ChatGPT and DallE may seem, or how useful they are in the day-to-day, their outputs are still a selection of most probably sequences and don’t contain any actual “thought.”

Impact

How Will Businesses Continue To Respond To Generative AI?

Those capable of competing with their LLMs will undoubtedly explore this as an option, but I want to talk about businesses adopting or resisting AI.

I’m quite close to the publishing industry from my experience at Forbes to my network, and having led tech for various Google publications. Publishing is one of the areas most affected by and most reactive to generative AI. For some of the publications I have contacts at I am aware of blanket policies now in place that forbid the use of AI in any part of the writing process. But if T9 texting from the early 90s could be considered AI, then it would be virtually impossible to adhere to such a broad requirement.

These policies for forbidding writers to leverage AI in favor of staying “human” exist due to a lack of understanding about what the technology is, how its sourced, and users being educated about it.

As LLMs evolve, companies will differentiate by adopting generative strategies and augmentation natively or by rejecting them altogether. There is room for success in both, but the latter fails to realize that AI in its current state is truly just a tool and not an “intelligence.”

My concern for businesses not competing against other AI companies however is that the pushback against AI-generated content will prevent them from leveraging AI on other parts of how they work. Even if there are ethical concerns about using ChatGPT to edit a blog post, its assistance in project planning, email writing, document summarization, and other areas is extremely high value.

You, Today

How You Can Leverage LLMs at Work, Today.

You can work better using AI today. Below are some tools and suggestions for how you can make this happen as a business leader without having to get your actual business to adopt them.

1. Feedback from Grammarly GO - Grammarly Technical is already an AI tool in terms of how it helps with grammar and spelling. Grammarly GO however is a fantastic resource in Beta that allows for getting quick feedback on your writing, tone, and clarity. Installing it on your computer will quickly help you be more clear and understand gaps in your blog posts, emails, presentations, and more.

2. Document Outlines with ChatGPT - This is the Swiss army knife of tools, and it’s not just for content generation. Much of what we do is templated or systemic. If you need to write out a Technical Design Doc, PRD, or any other document, you will find giving ChatGPT a prompt to write it for you and some details will give you the scaffolding of a document you can then edit and adjust on your own.

3. Text Expansion with ChatGPT - My favorite business use for ChatGPT is expanding on my text. Let’s look at this one in action.
   
   If you have to write a pretty standard follow-up to your team, you still need to be professional in the email you send. With an LLM you can turn this prompt into this email:

Further Reading

Books

Rebooting AI - If you consider yourself practiced in AI and you know the landscape you can skip this one, but if you need some more orientation, Rebooting AI concisely covers a wide array of topics, applications, markets, problems, and future directions we need to explore to get to the next level of AI integration, trust, and equality.

Blogs

• Using Machine Learning To Make Art