Here we are, perhaps at the last article of the year! 🥂
It's the perfect time to take stock of our use of generative artificial intelligence. This year, we've explored new possibilities, overcome obstacles, and learned to make the most of a tool that's revolutionizing how we work and think. But how can we best prepare for the upcoming year?
Here's a collection of insights and advice, seasoned with practical examples, to approach AI with awareness and a healthy dose of curiosity.
Use AI when you're more expert than it is
AI is powerful, but not omniscient. Think about it: would you let a self-driving car speed along without occasionally taking control of the wheel? If you're not well-versed in your work environment or subject matter, you risk making wrong decisions based on unreliable information.
Current models are extremely sophisticated but opaque. Without extensive experience in using them, you risk "falling asleep at the wheel" and letting AI provide you with information without verification or certainty about your working environment.
However, if you're more experienced, you can guide it toward the desired outcome, verifying the quality of its outputs step by step. You'll have a clear understanding of how it can help you and know how to steer it toward your goal.
An example? Imagine asking it:
"I want to create a startup on entanglement. Define the next steps and a use case to work on. How much money do I need for the MVP?"
The risk? Receiving seemingly sensible answers that are completely unfounded. Verification is mandatory: trust is good, but control is essential.
The moral: you must ALWAYS verify the quality of the output. But what happens if you're not able to do so?
Don't ask it for things that put you at risk if you don't know how to use it
In the previous example, a prompt with very little context generates different responses each time, with unclear logic:
ChatGPT might estimate 500-600k in necessary capital.
Claude might suggest 2 million for 24 months.
Would you trust that?
If you're not an expert, probably not. But generative AI is a formidable learning opportunity that will allow you to learn many things (if you know how to use it well). In my workshops, I often talk about seeing AI as an intern. What happens with human interns or students? They can use it as a tutor, relying on its knowledge or working in a learning environment that reduces risks from hallucinations or the production of risky content.
Perfect for summarizing... with judgment
Facing a long document? AI is excellent for saving time. But beware: its attention mechanism, however sophisticated, can overlook crucial details.
A real case: while analyzing an innovation grant, I noticed that the application deadline had already passed. The AI? It hadn't even flagged it. So yes to summaries, but with careful review on your part.
Generative AIs have a wonderful attention mechanism, but it's not always effective. You can use it to summarize large content, but for important decisions, personally verify the key points.
Its true talent: adapting and transforming
One of generative AI's strengths is its ability to transform content:
Expand or reduce text.
Adapt it for different audiences.
Create new formats: slides, videos, websites, music.
Don't hesitate to ask it. Generative AIs have "generation" as their core mission. They're imperfect and subject to hallucinations, but they know how to connect the dots. If you need 10 ideas, just ask. If you don't like them, you can ask for 20, 40, 100 more... It will keep responding, never exhausted. It will never judge you.
It has changed my life when I need to be "tickled" with ideas for titles, images, phrases. I usually create something of my own based on the ideas received.
Work where it can truly help you
Don't ask it for things outside its expertise. An example? Don't entrust it with complex mathematical calculations: that's not its job. This is why many chatbots have been supplemented with additional tools to fill the gaps.
It's better to use it to explore risks, opportunities, and scenarios.
For example:
`"What are the financial, legal, and economic risks if I sign this contract?"
Evaluate things from a risk manager's perspective: what issues might we face when starting to work with this supplier?`
This often takes me out of my comfort zone: If I'm not aware of certain risks, if I haven't studied a subject enough, if I don't have enough experience, I won't always be able to grasp the 'salient' and potentially dangerous aspects. Being told I'm running a risk helps me activate myself to verify what it is, explore it further, and reassess if I have the same vision.
Use it to produce content that requires high effort for low value
There's that famous "content that no one will ever read," inevitable in many work routines, that absorbs time and resources without generating significant impact. AI can be a valuable ally in handling these activities, reducing necessary effort and freeing up precious time for more strategic work.
Think about reports, meeting minutes, or those essential but unengaging bureaucratic documents. Let AI help you manage them. And the saved time? You decide how to invest it: it could become an opportunity to focus on higher-value projects or to recharge your energies.
Use it after completing your work
Reach your conclusions, put maximum effort into your work, and only then ask AI to tackle the same task. This approach is ideal for situations where quality is crucial and you're looking for alternative insights or confirmation.
By comparing results, you might discover whether AI reaches your same conclusions or, surprisingly, proposes paths you never considered.
An example? After writing this article, I provided GPT with my initial notes and asked it to arrange the text according to my tone of voice. Subsequently, I reread the result and returned to my version, perfecting some passages and integrating the ideas it had suggested.
Use it when you have something to do, not to do something
Except for the initial "tinkering" period where your goal will be to understand how a model works, how far it goes, what potential it offers (remembering to do it in areas where you're expert), always think about what your objective is with the conversation or activity you're doing.
Generative AI has the innate power to take you for a ride on its 'way of thinking,' asking you to do things you hadn't planned, and, I guarantee, will often make you waste time.
I'm not saying that serendipity shouldn't be welcomed as a possibility, but unless you're in a research phase, you need AI to achieve your ends.
Give it a clear role and make it reason before getting to the point
Context is fundamental for everyone, and it's determined by your prompts, its responses, and the documents you bring into the conversation. An initial exploratory phase, where you might ask which experts could help you perform a certain activity (or, if you already know, indicating them yourself), will allow you to significantly broaden perspectives.
The possibilities are almost infinite: you could, for example, involve experts who read your blog or integrate other knowledge sources relevant to the topic.
Don't trust blindly
Treat AI like an intern: it has disproportionate self-esteem and often seems to know everything, even when it's wrong. But unlike a human intern, it doesn't admit when it doesn't know something. Be critical and ready to correct it. If it were a real person, we might find it a bit annoying.
Have the courage to change the conversation
Sometimes conversations with AI get stuck. You might persist in trying to get it to deliver desired results, but it refuses. This can happen due to wrong contexts or compromised information.
The solution? Start fresh in a new conversation, better explaining the scope and situation, and starting from a draft that satisfied you.
So what?
Generative AI is an extraordinary ally, but it requires awareness and critical thinking. It's not a magic wand, but a tool to be leveraged to achieve your objectives. Remember: next year will be full of new challenges and opportunities. Use it with intelligence, curiosity, and a dash of audacity. But always responsibly!
See you next time! 🎉
Comments