Thought Stopper: AI as “Efficiency Trap”

TL:DR // Key Insights:?

• ???????? ????????? (Generative AI tools may distract us from solving the right problems because AI is making specifically trivial tasks more efficient)
• ???? ?????? ?????? ???????? ?? ???? ???????? ?????? (Diverse teams, even virtual ones with AI co-pilots, can generate better ideas than non-diverse ones)

The growing trend around Generative AI has brought about a myriad of nifty tools and solutions which are predominantly pointing towards “making our lives easier” by increasing the efficiency of mundane, automatable tasks, be it email writing, content generation, assisting with our research, helping us write routine code or connecting and cleaning our data for us. Overall, it very much feels like everyone is prematurely jumping on adopting solutions to pseudo-problems we have created in the first place by focusing all of our attention on pushing ourselves deeper down that famous “efficiency trap”.

AI undoubtedly has the potential to transform the way we work. But the majority of people seems to have been completely taken by the current hype around LLMs, specifically when it comes to quickly producing small pieces of “human-like” output. This has led to the adoption of any and all sorts of productivity tools and hacks - making the wheel in the solution space of work, meaning on the right side of the above depicted standard “Design Thinking Process”, suddenly spin much faster.

But wait, what’s the problem with increasing the productivity of creating solutions? Generally nothing, if your objective is just that, but then you might have already fallen for the above described efficiency trap yourself, and collectively we may be completely missing the point when it comes to paying attention to the actually burning questions of our times…???

What is truly lying at the heart of this issue is however a fundamental lack of focus when it comes to the first part of the process, referring to gaining a deep enough understanding of “the problem”, or what’s even more important, being completely sure that we are actually “solving the right problem” in the first place. When we are trying to understand how this issue is connected to the way we interact with our new AI technology, we immediately realise that in our interaction with the ChatGPTs, Bing AIs and Midjourneys of this world, problem finding, not problem solving should be our core focus. Meaning we should not be looking for AI to deliver us answers, but rather to improve our questions.?

This brings us directly to a new way of approaching any AI-enhanced process of working: Rather than obsessing over becoming the perfect “AI Whisperer” by perfecting our prompts in a way that AI is simply “doing our work for us”, which it can’t anyway unless you value your own creativity and ingenuity less than that of a pre-trained LLM, you should be focusing on collaborating with AI in a way so you can iteratively improve the respective “counter prompt”, meaning that AI will help you arrive at an ever increasing intellectual level of asking the most relevant and practically profound questions. In order to help increase the respective ability to ask the right question, I have set up a problem solving tool aiming at delivering back an improved “counter prompt” to any initial prompt, based on the concept of root cause analysis.

Human-AI-collaboration as “iterative problem finding”: www.counterprompt.ai

Applying such a “problem finding approach” might actually be an effective way to address what has been described as the main “barrier to innovation in generative AI”, namely its still prevailing lack of vision. Because it is only when we start to think about bigger use cases for AI that we will do it justice as the technological progress that is “more profound than electricity and fire”.?

When it comes to increasing our understanding of complex, multidimensional problems, there are actually a whole series of tools we can use for exploration, analysis and insight generation. One might be to simply map all existing dimensions of that problem onto a radar chart in order to better understand its overall dimensionality as well as our own ability to fully assess the respective complexity within each dimension. Another, more advanced option might be to set up? an interactive way of analysing the relationship between each dimension by using knowledge graph technology in association with AI, for example via tools like schema.org.

Another great way of untangling multidimensional complexities is linked to the collective analysis and brainstorming of ideas with the aim of jointly discovering also non-obvious solutions, for example by iteratively mapping all existing knowledge gaps and intellectual white spots on our collective knowledge map, something that can be done on many levels, including mapping the most used keywords or most prominent businesses involved within patenting.

What we can discover during this process is specifically the value of inherently diverse teams when it comes to arriving at the best ideas. Interestingly, this also holds true when we work in entirely virtual teams, consisting only of ourselves and our co-piloting AI. When testing this hypothesis, we achieved on average an increase of 9% in ideation quality when running the otherwise exactly same task on an AI model that was prompted to act as if it represented a diverse, multidisciplinary team of experts, rather than just being asked to come up with non-obvious, original ideas across various different contexts.

Plain / non-diverse prompt asking for?“original ideas”

Prompt asking to simulate a diverse team to come up with “original ideas”

This might leave us with the following thought: The best us of AI might be one that challenges our thinking in order to help us think further rather than replacing our own thinking to the point that we will even rely on robots to do the most mundane of tasks, like getting (un)dressed…