A few years ago my thought were on the terminator…. My thoughts suggested that computer a bit less “stupid”. I still believe machines have more functionalities that mimics some behaviours humans believes exhibit some level of intelligence.
Smart may not be so smart
In the last 6 years many devices have become smart. Watches, lawnmowers, heating systems, are just part of the the list. This adjective gives the impression some devices have some level of intelligence or some acumen. Instead this adjective is suggesting a system made of components connecting using the Internet. The Internet of Things (IoT) hide many processing units and the human interaction is just the tip of the iceberg. The hidden layer is likely to be part of some data centres with high computational resources. The users are oblivious of these hidden layers and any network communication these systems rely on. So, instead of having a swarm of robots trying to move forward in one direction, we have a swarm of devices; each device has a specific purpose.
Smart may not be so smart. We connect devices to the Internet, a network of networks. Data are processed using a variety of algorithms. The results are communicated back to our devices. It is a smart integration of systems, but not any demonstration of intelligence by the machine. The individuals involved their development demonstrated thought many types of intelligences; innovation, problem solving, creativity, exhibition of emotions.
Intelligence and learning,
“It is not that I’m so smart. But I stay with the questions much longer.” ―Albert Einstein
“Education is no substitute for intelligence.”― Frank Herbert
Intelligence is an idea. Some individuals perceive it as the ability to learn new skills. Some other to solve problems. Some other it is being creative. Various community of research define intelligence in their own way. Many tests exists to measure our intelligence
Multiple perceptions of intelligence in machines
Artificial intelligence (AI) and machine learning (ML) is affected by the same phenomenon. It is amusing that artificial intelligence was coined and invented to simulate intelligence and study it without having access to individuals. Science fiction has portrayed AI in optimistic manner. The key point is to acknowledge different people will perceive any ideas from their own interest and perceptions. Conscious intellectual activity can develop those and make us grow through our life.
Turing test is based on natural conversation between machine and human, that is indistinguishable. The probability of repeating patterns of words in our day to day conversation has made this possible. In this previous work in deciphering German messages, it is believed the assumption that every message had a famous salute may have led to decoding messages. In some point, machine needs human understanding to a problem domain to compute their own assumptions.
Perception of predictive algorithms
IT specialists perceive these techniques as some element to integrate in some bigger and greater computing system to increase the production of information. The latter can only inform humans of unknown findings. The fact is it is happening, but the ML is predicting correctly the price of houses in areas where it is know it is likely to be expensive or more affordable, for example. It is a basic inference that human are more efficient to compute with experience. The ML may be processing quicker large information to indication some predictions …
Computer scientists may view ML and AI as some algorithms we can apply on some datasets to obtain some outcome. The techniques can be tricky to understand and therefore some guerrilla stance can lead not considering other techniques that may be more suitable. Computer scientists with some statistical and probability knowledge starts to decompose the algorithms and learn the type of data each technique may work. These individuals have become data analysts and data scientists. Computer scientists with strong mathematical background are likely to model first and use formal methods to validate their findings and share with the community.
In my experience, statisticians can sometimes struggle with the idea of AI and ML. They work is to use statistical methodologies to analyse some data to explore the extend of some association; they findings is always based on likelihood and significance. There always is a shroud of doubt. Statisticians looks at the data first, and then carefully decide the statistical method to apply. It is unlikely a linear regression is applied to some datasets demonstrating some polynomial characteristics or clustering information. Statisticians have yet to find the information of suitability for AI and ML algorithms. An increase use of advanced statistical methodologies and approaches is increasingly being brought to medical data science. The outcome of such activities can lead to life and death.
ML rely on statistical methodologies and probabilistic models hidden away algorithms and brought to a general audience as a magic wand. Unlike other software packages, mathematical knowledge is necessary to fully appreciate the techniques. Optimisation, calculus, solving linear equations, graph theories, statistics, probabilities are important. Problem solving and analytical skills are also most relevant along side some specific domain knowledge of area ML and AI is being applied. The Intelligence Augmentation has been brought to us as a revival of ML. I feel statisticians have augmented our intelligence for years, so perhaps we need to continue to improve our cognition and learn from these individuals in this endeavour. No more short cuts.
It is pleasing image recognition and translation has been successful in predicting the potential of medical conditions at an earlier stage, then the brain. The speed of computation is working well alongside the specialists. It is a niche market in which such applications are suitable. On the other hand, this impressive speed of computation has a dark side. Malicious attacks and ways to negatively affect our increasing reliance of large distributed systems can be fuelled by AI and ML. These individuals must be outstanding computer scientists as well as mathematicians. It would be much better if their energies could be spend on protection, rather than war fare ….
Can AI and ML solve problem?
In one of my most recent post the limitations of data science were discussed. Machine could assist in processing more effectively combinations of operations to solve a problems. More information can be found in my Doctoral thesis.
Interesting Computing 