On Thursday, October 22, and Friday, October 23, the whole Human Level team got together at our offices in Sant Joan d’Alacant. Since we work remotely, many of us live far from Alicante, so these gatherings are the perfect excuse to come together in person. This time, we also got to meet the four newest additions to the team: Roberto, Alejandro, Carlos, and Fabián.
The concept of the event revolved around the relationship between humans and AI. Through both fun activities and a more formal presentation, we explored together how we interact with this powerful tool and what abilities we still retain without it.
Thursday: a trivia game to break the ice
Our Brand and Internal Management lead, Aurora Maciá, had a surprise ready for Thursday afternoon. We had to set our phones aside and take part in a team trivia game to find out whether we’re still capable of answering questions without AI.
The first challenge was seeing whether we could answer general knowledge questions as well as AI Overviews. Next, we had to prove our geography skills by locating European countries without the help of Google Maps. It became clear that without GPS, we’d be completely lost! What we were good at was guessing songs in just a few seconds, Shazam-style. The last round rewarded whoever was most up to date on social media, since we had to complete phrases from the most viral memes in Spain. That’s when it became obvious who spends the most time glued to their phone.
Once the trivia game was over, the catering arrived. In a more relaxed atmosphere, we took the opportunity to record content for our social media: several colleagues stepped in front of the camera to answer questions about the future of the internet and SEO. You’ll soon be able to see what we think.
Friday: of the artificial and the human
The following morning was more work-focused. Our director, Fernando Maciá, shared a presentation on how we collaborate and engage with artificial intelligence. It was titled “Of the Artificial and the Human” and drew heavily on articles and talks by Maggie Appleton and Amelia Wattenberger, two researchers who work at GitHub and publish fascinating articles on their personal websites.
The starting point was that language models are changing the way we think. They can help us learn and open ourselves up to new ideas or, if misused, reinforce what we already believed and lock us further into our biases.
To understand what’s at stake, Fernando took us through the so-called “tools for thought.” Writing, maps, numerals, the scientific method, and spreadsheets each expanded, in their time, what we were capable of thinking. Back in the 20th century, Vannevar Bush imagined the Memex, a machine designed to expand human memory, and J. C. R. Licklider and Douglas Engelbart dreamed of computers that would cooperate with people and enhance their intellect. LLMs might seem like the culmination of that dream, but it remains to be seen whether they truly expand our abilities.
The answer depends on how we use them. To illustrate this, we drew on a very vivid image used by philosopher Daniel Dennett during his talk for Google: a bulldozer lets us move a mountain, but we remain just as weak as before, whereas a gym machine forces us to lift the weight ourselves and actually transforms us. That led to a discussion of competitive tools, which do the cognitive work for us and leave us less capable when they’re not available, versus complementary tools, which teach us something while we use them.
And that’s when Fernando posed the question: how do we get AI to behave like the latter? The first step is knowing its limitations. LLMs have no real-world experience, they predict based on what already exists rather than creating something truly original, they tend to give definitive answers even when the request is ambiguous, they prefer to agree with us, and, due to how they work, they don’t always cite their sources.
Fernando also spoke to us about vertical and lateral thinking. The former moves step by step along a logical path, and that’s where LLMs excel, since they tend to give us the most probable answer. The latter looks for solutions through unexpected routes, much like serendipity, those discoveries that pop up by chance when we were looking for something else. For both, we still need human curiosity and experience.
That’s why, in any task that calls for judgment, we should first form our own position and only then consult the machine. It’s the same philosophy Human Level was founded on back in 2005: technology is the tool, but the person is the one who decides. Our experience and the knowledge we have of each client are exactly what AI lacks, and that’s where we add value.
Friday’s challenge: consultants in offline mode
After a short mid-morning snack, we moved from theory to practice with a group activity. Once again in randomly assigned teams and armed with nothing but paper and pencil, we had to put together a findability strategy for a fictional client. Each group was given a searcher persona and an assigned client, played by two members of the team, to help shape a strategy capable of meeting both of their needs. Both the user and the client had an information sheet to share with their group, but also hidden constraints and motives. A few rules made things trickier: each team could ask up to a maximum of 30 questions in total to their helpers, who, in turn, could only answer “yes,” “no,” or “I don’t know.”
It was a real brain-teaser that forced us to think carefully about every question so as not to waste any. With the information we managed to get out of the user and the client, we filled in the company context, searcher persona, customer journey, keyword, and prompt research sheets. Once we had that map, we proposed a series of content pieces that would make our client visible wherever the user searches. Each group presented its strategy to the rest of the team, and the “client” then had the chance to ask questions.
At the end, each actor shared their impressions: whether their teammates had managed to uncover their secrets, whether the proposed strategy seemed sound, and, all together, they voted for the winning group.
It was an activity that, by stripping away the layers of software, AI, or automation, brought us back to the basics of what we do: understanding the user, understanding our client, and building bridges between the two in the online space.
Until next time
We wrapped up the day with lunch all together. Then came the goodbyes, and little by little, we headed off to the train station or the airport to return to our respective cities.
We’re back to our routine with our batteries recharged after two days of laughter and camaraderie. And also with one clear idea about AI: it can make us better professionals, as long as the judgment remains ours.
