Self-driving cars do not yet roam the streets. AI has yet to wait to independently make a computer game or a larger part of software, and chatbot assistants are only carefully deployed with human assistants as backup. We continuously wonder how much longer it will take until AI reaches human skill level in these tasks - or, erstwhile does AI become "truly" intelligent. This question frequently poses an implicit presumption that human intelligence is the perfect plan to which AI should aspire to. However, we request to remember that at this very minute various forms of AI environment us which overshadow human abilities in their respective areas of operation.
In this article I share any intuitive arguments and examples of limitations of human cognition, compared to the abilities of current digital systems. I aim to show that while human performance should be pursued as a gold standard in any tasks, in another tasks even human experts can be easy beaten by an inhuman AI. This is due to the fact that we have been subject to evolution – a purposeful process which honed a selected set of skills to handle a circumstantial set of challenges posed by our environment. However, the modern planet of large and frequently abstract data streams creates previously unknown challenges. To solve them, we request a different kind of intelligence than our own. Here we arrive at the word “inhuman AI", which raises alternatively negative connotations. It is usually associated with catastrophic visions of powerful, evil AI hurting humanity. However, in practice we already exploit the inhuman side of AI to our large advantage.
Where humans underperform
Historically, we tended to measure certain tasks as easy due to the fact that most healthy people can execute them without much conscious effort. This applies to for example walking, language knowing and image analysis, specified as segmenting objects or assessing depth and distance relations in a picture. another tasks utilized to be thought of as hard due to the fact that only selected (most talented or hard working) people could execute them. In the early days of AI this notion applied to e.g., games specified as chess and go. evidently AI has beaten human planet champions in both these games (DeepBlue and AlphaGo/AlphaZero), as well as in online RTS gaming (OpenAI’s Dota player system) and poker (CMU’s Texas Hold’Em player system), among others. At the same time, natural language processing and imagination systems are inactive far from human performance, and brittle in terms of generalization abilities.
The head is simply a strategy of organs of computation, designed by natural selection to solve the kinds of problems our ancestors faced in their foraging way of life, in particular, knowing and outmaneuvering objects, animals, plants, and another people. – Steven Pinker, "How the head Works”In fact, the most “basic” human or animal abilities specified as walking on 2 or 4 legs or tiny motor skills of the hand are characterized by considerable computational complexity. They are so hard that no artificial strategy can match them – yet humans can execute them (almost) subconsciously and effortlessly. This is an effect of evolution working tirelessly for millions of years to refine these abilities to perfection. In comparison, higher cognitive abilities specified as complex logical reasoning and abstract reasoning are an early invention in terms of evolutionary time scales – they started to make only around 100,000 years ago. Their early phase of improvement manifests in the way advanced cognitive skills seem to be harder and require far more deliberation and conscious effort than for example walking. Interestingly, “primitive” and “advanced” human neural skills appear to have conflicting aspects. advanced cognitive functions have developed on top of more primitive intellectual structures, so the resulting performance is not entirely optimal due to erstwhile "design choices". Possibly, this is connected to multiple cognitive biases of the human head which lead us to making wrong, irrational decisions [1,2,3]. Examples of specified biases are the confirmation bias (the tendency to interpret and remember information in a way that confirms our beliefs, views, or expectations), or the availability bias (the tendency to erroneously delegate greater importance to events only due to the fact that they were seen recently), among many others. 1 of the explanations for the existence of cognitive biases is that the human brain simply lacks the power to process specified complex situations and must hotel to shortcuts or severe simplifications [4]. Unfortunately, as humans we are not aware of the biases and are usually convinced of acting rationally and objectively.
Thus, it is evident that in the past we were many times incorrect in our assessment of “objective” task difficulty, as well as our own inherent cognitive limitations.
Games – a request for more memory
It is actual that chess and go require complex strategical thinking. However, the example of the most successful AI game playing algorithms shows us that a vital component is the ability to analyse game trees – that is, the ability to look profoundly into the future. The game tree is simply a representation of combinations of future decision sequences. Consider a chess board during play:
From here, multiple moves are possible – for example, we can decision the Black Queen to A3, B3, or D4. Each of our moves can either deteriorate or enhance our position on the board and has its place within a larger strategical scheme. However, we must besides consider what the opponent will do after our move. Each possible opponent reaction opens up multiple fresh movement opportunities. The resulting game tree of all our possible moves and all opponent possible moves grows to a immense size:
In order to realize the volumes of data we are dealing with, let’s consider any statistical properties of chess and Go, specified as the branching factor, average game length, and game tree complexity. The branching origin expresses the average number of legal moves available to the player each turn. Chess has a branching origin of 35 and Go – 250. A single chess game ends in 70 moves on average, while Go - in 150. What interests us the most is game-tree complexity, which denotes the number of leaf nodes in a full game tree. For chess this equals 10^123 and for Go – 10^360. These numbers are actually hard to comprehend. The number of atoms in the universe is “just” 10^87.
For an AI system, the amount of available memory and compute is adequate to rapidly produce and store game trees of large size (usually not full trees, but inactive large), impossible to remember for humans. Modern game-playing AI systems can selectively “look through” many possible game strategies in order to reason about best possible moves and explicitly decide which paths are viable of consideration, and which should be dropped. This allows for selecting the best optimal strategy in nonsubjective terms, even if it is not the most apparent one. In fact, a peculiar AI-inspired word has been coined to denote weird, unexpected chess moves – they are known as "computer moves”. This refers to how AI-generated moves frequently look confusing or even senseless at first – yet their utility is revealed much later in the game.
As a result, human masters have consistently been losing chess plays to AI within last 15 years, and it is universally agreed that no human can possibly win with a high-grade AI chess player.
Abstract numeric problems
As noted before, humans are evolutionarily primed to swiftly process language and imagination signals. We are way worse at processing symbolic data, peculiarly numbers. However, many problems nowadays are expressed in an inherently numeric format - imagine signals from sensors in mill production lines, financial data, or signals from medical devices recording various aspects of the heart rate, oxygen saturation, perspiration rate, and many others [5] in real time. They form sequences of natural numbers, in which patterns request to be spotted in order to derive any useful information. An highly crucial ability is for example the skill to place which numeric datapoints are closer ("more similar") to another datapoints. This way we can detect akin cases of illness or mill malfunction to the ones which happened in the past.
Let’s see how well we as humans can execute pattern spotting in natural numeric data. Consider a bunch of point coordinates in a 2D space – each row denotes a pair of X and Y coordinates of a single point. Think what popular form these coordinates could represent:
[-6.13, 6.09],
[ 5.57, -10.06],
[ 0.005, -4.71],
[ 2.18, 7.24],
[ 5.99, -1.14],
[ 4.27, -3.59],
[ 6.36, 1.19],
[-9.16, 7.92],
[-4.76, 6.99],
[-9.14, -3.50],
[-6.27, 5.98],
[1.94, -11.00],
[ 6.22, 2.31],
[ 11.85, 5.32],
[ 1.27, -4.81],
[ 0.39, -11.02],
[-6.15, 6.08],
[-8.49, -5.24],
[ 5.85, -1.48],
[ 11.69, -3.71],
[ 4.08, 13.22],
[ 1.43, 7.53],
[ 2.69, -10.91],
[ 11.92, 5.12],
[ 11.01, 7.18],
[ 3.96, -3.84],
[ 1.83, -4.74],
[ 6.31, 0.19],
[ 11.85, 5.32],
[ 4.51, -3.39],
[-6.39, 5.88],
[ 5.31, 12.67],
[-9.27, -2.99],
[ 2.64, -10.92],
[ 1.28, 7.58],
[ 2.08, -10.99],
[ 12.58, 0.15],
[ 0.84, -4.81],
[ 8.34, -8.32]
Judging by the numbers alone (without an effort at plotting the datapoints), it is virtually impossible to say which form they form, or or what are the distances of points along the line which is drawn.
Only after plotting does a pattern start to emerge:
And gets more pronounced with more points:
This is where the area of data visualization takes its roots: it is far, far easier to observe a tendency in visualized data, than in pure numeric form. If we could have said that the pattern is simply a spiral judging from natural numbers alone, the large data visualization business with companies specified as Tableau, Grafana or Dataiku would never have been born.
Now, let’s take a look at an algorithm called tSNE [6], a dimensionality simplification method. This algorithm can “understand” data by ingesting a pure numeric form and based on this alone, it can reason about the “true” proximity of points on complex nonlinear manifolds very accurately. The spiral example is no large deal for tSNE, which correctly models the distance problem on the "Swiss roll" form (marked with the solid line):
Note that the 2D example we have analyzed so far is in fact an highly easy case, since most numeric signals would have hundreds or thousands of dimensions. Now, it is not even possible to visualize a hundred/thousand-dimensional space for human inspection, or even imagine specified a space. Here we must trust on machines fully. In fact, tSNE and many another device learning are designed to handle thousand-dimensional spaces.
Enter large Data
Our final example will mention to large data – massive datasets composed of billions of data points. Many vital areas of human life trust on large data processing:
- Fuel and stock optimization tools for the transportation industry,
- Live road mapping and way planning for both human drivers and autonomous vehicles,
- Traffic modeling, prediction of traffic jams and accidents,
- Meteorology: weather prediction, disaster prediction, survey of global warming from massive amounts of signals from satellites and sensors,
- User request prediction in e-commerce, online music and streaming services,
- Monitoring wellness conditions through data from wearables,
- Discovering consumer buying habits derived from millions of users of sites specified as Amazon,
- …. and many more.
From its very definition, large data modeling aims to discover insights hidden in tremendous amounts of data. In order to detect patterns, oftentimes it is essential to perceive a dataset as a whole. tiny chunks of data frequently do not disclose any meaningful pattern and can besides contain false patterns which vanish on a large scale. Consider a patch of an LCD screen and a full screen:
While utilizing displays and screens, we request to keep a certain distance to perceive the full picture. Seeing only tiny pieces of data (such as a fewer pixels), it is virtually impossible to say what the screen displays. It’s even worse with real numeric large data, due to the fact that we have no prior intuitions about what certain signals could mean.
Imagine an army of agents with limited capabilities of memory and processing power, able to look for local patterns only. This army is by no means guaranteed to find a legitimate global pattern. The full dataset must be processed by a powerful agent aware of all its constituent parts. (In fact, there is simply a branch of discipline dedicated to studying whether global tasks are solvable by systems of limited agents, called Global To Local Theory.)
Even if we tried to experimentally verify the hypothesis that our army of humans can discover global patterns in large data, we immediately run into applicable problems. Due to slow and tiny working and long-term memory, as well as slow process of data transmission (humans request to read or hear information, which is painfully slow compared to digital data transmission), it is impossible to anticipate that human agents can process billions of data events in reasonable time and with reasonable effort.
Summary
As we can see, quite a few vital applications of AI trust on fundamentally inhuman skills. However, the inhumanity of AI can sound like a negative trait, or even threatening erstwhile perceived from an ethical point of view. Yet it is evident that this kind of computer intelligence is needed to patch up any inherent flaws of our cognition. Inhuman AI pushes the possible of our civilization far beyond our evolutionary design. With amazing achievements specified as GPT-3, DALL-E, Imagen, and others trying to catch up with human skills, let’s remember that we have been already beaten many times before, yet at different tasks – and to our advantage.
Bibliography
[1] Korteling, J. E., Brouwer, A. M., and Toet, A. (2018a). A neural network framework for cognitive bias. Front. Psychol. 9, 1561. doi:10.3389/fpsyg.2018.01561
[2] Korteling, J. E., van de Boer-Visschedijk, G. C., Boswinkel, R. A., and Boonekamp, R. C. (2018b). Effecten van de inzet van Non-Human Intelligent Collaborators op Opleiding and Training [V1719]. Report TNO 2018 R11654. Soesterberg: TNO defence safety and security, Soesterberg, Netherlands: TNO, Soesterberg.
[3] Moravec, H. (1988). Mind children. Cambridge, MA, United States: Harvard University Press.
[4] Friedrich, James (1993), "Primary mistake detection and minimization (PEDMIN) strategies in social cognition: a reinterpretation of confirmation bias phenomena", intellectual Review, 100 (2): 298–319, doi:10.1037/0033-295X.100.2.298, ISSN 0033-295X, PMID 8483985
[5] Annica Kristoffersson and Maria Linden (2020) “A Systematic Review on the usage of Wearable Body Sensors for wellness Monitoring: A Qualitative Synthesis” https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7085653/
[6] L.J.P. van der Maaten and G.E. Hinton. Visualizing High-Dimensional Data utilizing t-SNE. diary of device Learning investigation 9(Nov):2579–2605, 2008.
