Tricks to Try Out
Warning: Hazardous material here. Use responsibly.
The utopia of altering the code of our hardware, our "meat," for anything and everything has proven to be a bit geeky (it's amusing how it remains the most irreplaceable and impressive aspect even for the most advanced machine). However, we can strive to enhance our software, especially since it's a shame that such unique and extraordinary human beings are sometimes lacking in intelligence.
Understand the evaluator. When an agent seeks to be intelligent, it is initially quite intelligent to comprehend who is evaluating it. It's worth noting that this goes beyond what current artificial intelligence systems typically employ, which is optimizing for the goal set by the evaluator. We can delve deeper and strive to understand how that goal will be judged in itself and how to satisfy it. Let's consider an illustrative scenario where the intelligent agent's goal is simply to determine if it is sufficiently geeky to perform a senseless task disliked by the evaluator. In the case of human evaluators, this encompasses all relevant aspects of physiology, psychology, and sociology. (Take note, for instance, how ChatGPT appeared significantly smarter after humans taught it the type of content they genuinely appreciate, similar to what's already happening in recommendation algorithms on platforms like TikTok, YouTube, Instagram, and others.) Also, it's important to recognize that this can apply to the individual themselves if they act as their own evaluator. A more dystopian standpoint may even involve attempting to change the evaluator—whether altering the person themselves or literally replacing them with another. There is an entire realm of possibilities here, some of which may be unsettling. Let's see if the GPTs don't take notice of this.
Be unpredictable. If it is a context where, for example, the success of our intelligence depends on the opponent not being able to understand it. Note that in a relative context, such as a social context, even being stupid can be the smartest thing.
Experiment with symmetry operations. For example, the simple opposite of the thing. Even the opposite of a good idea can be a great idea. Note also that there are different opposites in relation to different dimensions. And there are other operations of symmetry beyond the mere opposite, such as translation, rotation, etc. Considering another example, Reductio ad Absurdum is a common method of proving a statement by assuming its opposite and then deriving an absurd or impossible conclusion. Note how it can also be used to find ironically ...good solutions.
Be counter-productive. Instead of looking for direct solutions to a problem, think about the actions that could make the situation worse. It can help to identify approaches that should be avoided.
Reverse induction. Starting with a solution and moving backwards step by step until we realise how we got from our problem to it.
Self-nudges. For example, always carry some subconscious induction cards, with what we want our subconscious to process so that then, "out of nowhere", interesting insights appear. This way we use the "confirmation bias" in our favour, where we start to find things that meet what we want. And they have to be "nudges" that we can't ignore, otherwise we don't even see them in everyday practice, or we get used to ignoring them.
Caching. For example, regularly reviewing the most important information to keep it fresh in your mind. This makes it easier to recall them when needed. Or having certain algorithms, like this list of tricks, always at hand.
Looking for analogies. Lateral thinking. Sometimes the solution already exists, just somewhere else and in a different guise.
Combinations. Most new "creations" are a mix of creations that already existed.
Live with paradoxes. Giving a little time to things that don't seem to make any sense.
Synchronicity. Sometimes you have to wait, work, get lucky, etc. for "the stars to align". Having the elements of a solution out of sync can result in not having the solution when we do.
Diesel thinking. An environmentally unfortunate analogy, we know, but sometimes there are mental processes that take time and aren't giving many conscious signals but are actually getting there hard. Being in too much of a hurry can mean simply ignoring all that potential. Even a poor computer when it's sitting there a bit slow makes the human user freak out straight away and want to stop it.
Simulations. It can be a computer simulation, in our heads, in other people's heads, etc.
Prepare possibilities. Proactively consider possible scenarios.
Do "scientific" experiments.
Sensitivity analyses. By changing each variable a little, which variable seems to be having the most effect? For example, we may be in situations where most of the consequences are due to a minority of the causes (as described by the Pareto Principle).
Map the space of dimensions to consider. For example, considering the amount of something, the time and the cost, we soon have a 3D space to explore with the various possibilities combined.
Follow the gradient. Like a dog following an algorithm that takes advantage of computations already happening outside, where it just has to follow the direction in which the odour is most intense.
Use noise. Great for getting out of sub-optimal holes. Sometimes too neat ways of storing or thinking about information don't lead to exploring other, more interesting possibilities.
Use randomness. Also great for getting out of sub-optimal holes and for arriving at solutions that straight logic would not easily reach.
Use your own seeds. In sufficiently complex systems, it can be extremely important what kind of starting position you start from. This can mean that even someone very stupid can beat a lot of smart people simply by using the starting position they have chosen. At the limit, a solution may have been found by someone and not even be possible for anyone else, from another position, to reach it. Note that to be intelligent you have to spend time, resources, computing, etc., all limited things to "get there". There are people who by some stroke or other of the universe may "already be there".
Use natural selection. Many and insistent iterations of two phases: diversity and selection.
Delegate. Delegate to tomorrow, to someone else, to the computer, etc.
Bound the problem. Take it to the extremes and then what we're interested in at least should be within that - that's not bad.
Trapping. Once you have the territory defined, narrow it down further and further to where the problem or solution actually lies.
Moments without distractions.
Moments with distractions.
Abstracting. There are always several levels of "computational depth" at which we can think about a problem. It can be at a high level, like icons on your mobile phone screen, it can be at a low level, like the computer code that controls every pixel of every such icon. For example, often creating a map, or symbols, or colours, that more abstractly represent something can exponentiate our ability to think about it.
Think in networks. Organising ideas into elements and relationships between elements. Structures emerge that often even seem to "think by themselves". It can be done on paper, in digital format, etc. Note that much of the success of our brains and the famous artificial intelligence is actually "integrated memory", i.e. putting things into a necessarily interconnected format that cannot simply be separate boxes of junk. Just doing this process can make the solution flow naturally afterwards.
Feedback. A living organism without feedback dies. And there are various types of feedback, as well as various speeds.
Feedforward. For example, before an important presentation, a person may request feedforward from colleagues or mentors, seeking guidance and suggestions to improve their performance before the event itself.
Compouding. Thinking on top of what has been thought, and so on.
Collaborate. For example, a whole human being can emerge from the collaboration of cells. There can be collaboration in terms of people, data, goals, etc.
Competing. To some extent, competition is a type of collaboration. Many things when they are separate do not contribute as much to solving the problem than when they are in direct confrontation. We use this trick in markets, why not use it to solve other concrete problems?
Reset. Think afresh. And to take care with each new step, so that we don't fall into an alley again.
Chunking. Segment the problem into mini-problems (and, if necessary, each of these into mini-problems, and so on).
Deconstruct. Often what is most problematic is the way we are looking at the problem. Getting to what it really is, fundamentally, and going from there.
Playing with distance. Distance in time, space, or any other variable. Not getting stuck in the perspective we are having here and now. And often the very movement of switching between one perspective and another holds the key.
Structure. Organise knowledge in a way that makes it easy to retrieve and apply. This may mean taking structured notes, creating mind maps or using mnemonic devices.
Process in parallel. Multi-tasking usually does not pay off, because of the trade-off between tasks. But actually doing several things at the same time can be possible and desirable. For example, when we drive and listen to a podcast, or when we write while the computer processes data.
Version Control. When faced with a new possibility, note the previous state and the new state, which is to be experienced. At the limit, if we don't like it, go back to the previous state.
Think out loud. Writing, talking, drawing, whatever. It's different to just think in your imagination and actually put it out there in a cycle where your perception is actually realising something that, even though it was made by you, has now come from outside.
Repeat. Sometimes doing something just once may not be representative. It may even be just a case of bad luck. In many cases, the more we do, the closer we get to the real value to be found.
Explain. Often just the exercise of explaining the exact problem to ChatGPT, to a colleague, to a child, alone leads to the solution.
Naive optimism. It helps to try what seemed impossible but turned out to be possible. (Or else to waste time, but anyway.)
Capitalise on peaks of intelligence. We are not always at our best, of course.
Gather the best insights in some sand-box-like intermediate medium. Then in a later phase analyze what you have in there.
Cross-test pairs of variables to assess correlations and false negative and false positive rates. In other words, use matrices of 2 rows (presence and absence of variable 1) and 2 columns (presence and absence of variable 2).
Do nothing. Do not overlook the possibility that the problem can be solved by itself.
Expand our working memory. The amount of things we can "hold" with our mind is ridiculously low. Use other strategies, from using other parts of our body (like a child using fingers to count, which can be seen as a "somatic" working memory) or the very position of our body in space, to using paper, computers, or other people.
Sleep, eat, drink, exercise. Remember that our nervous system is a living system. We need to cleanse it, rest it, eat the necessary nutrients, moisturise it, etc. A mobile phone that has fallen into the toilet or is boiling or has run out of battery, ouch, that's a big problem. We, who are slightly more complicated ...ah, never mind!
Avoid the "Mistakes to Avoid" (see the other list).