AI for Non-Techies
?
What is AI?
AI basically copies human capabilities and the way we perceive our environment with the potential to
at any given time.
?
Throughout the history of technological innovation, we have been mimicking existing models in nature. With evolving progress, we could take on more complex challenges and with the emerging field of AI we seem now to have set out to tackle the blueprint of the Homo Sapiens with its tremendous abilities. Contradictory to what we may think at first, this development is, although artificial, quite natural as technological evolution is simply a part of the natural human evolution. Therefore, it is no surprise that in today's digital world the ability to process huge volumes of information intelligently matches the dynamic principle of evolution - Survival of the fittest.
?
Even possible worries around (creating) a force that can compete with us or that sets the bar even far higher than we can reach appears to be natural. Evolution has always been fueled by ongoing competition and we have never been superior in everything. Every form of living comes with a trade-off. Some can fly, some can dive. The human species cannot see UV light, cannot see super far and doesn't have receptors for Sonar. Some snakes can see UV light, eagles can spot prey from kilometers away, bats and dolphins have built in Sonar/Echolocation and dogs can track scents over long distances and across various terrains, etc.
?
With developing technology, we could not only match abilities of other species, but we could even supersede them far beyond what we first thought was possible. We can see into space with telescopes, we navigate submarines with built in Sonar and we use night vision googles to see in the dark. Once technology has become part of our life, we don't even tag it as artificial anymore as it is just normal. AI simply seems to be a natural continuation of what we have been doing for some time already.
To mimic human capacities, AI puts together different building blocks. Same way as the human organism. A set of various systems and protagonists working together in an organized fashion, aiming to create an artificial form of life that can interact with its environment autonomously.
?
Machine Learning
o??? Supervised learning
o??? Unsupervised learning
o??? Reinforcement learning
o??? Deep learning / Neural networks
ML Models are various algorithms composed to coherent flows or networks that learn over time within a specific problem or perform a defined task. (Image recognition, logical thinking, conclusion making, playing complex games, learning from experience, reasoning about complex problems, …)
NL Models use various algorithms, Text to speech technology, grammatical rules and other resource such as dictionaries and databases. It can generate human language, answer questions, translate language, make conversation, receive and execute commands, etc.
Equipped with various sensors and the ability to interpret input enables AI to perceive the environment. (Vision, sound, touch, temperature, smell, …)
In the efforts to mimic and copy nature, robots have already got an astonishing level of human characteristics such as movements, reactions and gestures.
?
?
?
Where does intelligent software end and where does AI start?
Intelligent software can be seen as an intermediate step to AI. It can behave intelligently and even learn but it must be explicitly programmed for a certain scenario and task, and it is limited to a predefined framework. AI is thought to get to a whole other level that has a feel of an independent unit to it. It aims to replicate human intelligence and capability across all domains. It can learn from experience and autonomously interact with its environment. Intelligent software is stiff and static whereas AI is flexible and dynamic. As a rule of the thumb, you could say that something is more AI, the more human-like it feels.
o???Email spam filter
An email spam filter learns and improves by analyzing incoming emails, identifying patterns and characteristics associated with spam, and adjusting its filtering algorithms based on user feedback and new spamming techniques and evolving email threats.
o??? YouTube algorithm
It can recommend videos and content based on user interactions, viewing history, engagement metrics, and video content metadata (watch time, likes, number of comments, shares, subscriptions, watch history, trending and popular videos, key words in comments, key words in title, key words in description, tags, …). It does a good job in recommending videos by comparing a set of data in an intelligent way.
o???ChatGPT
ChatGPT is a text-based AI with which one can interact in a conversational fashion. It can outline the YouTube algorithm in great detail and even provide a written code of a similar one.
Please note. A distinction may not always be clear as AI technology is enhancing existing applications of intelligent software as we speak.
?
Some examples of AI technology - Get excited!
?
??
Where could this be going? (General AI)
AI is going to be a somewhat autonomous assistant that enters all aspects of civilization. From an individualized level as a personal assistant to a nationwide autonomous unit that manages basic infrastructure such transportation, goods supply and payment systems. It's going to be a link between the human collective and the artificial collective. Depending on the area of use and sales models, there will be assistants with various competencies, capacities, shapes and sizes. The appearance is going to be virtual, holographic or physical. Systems are going to be more dynamic and flexible due to real time data feed and the end-to-end connection of almost everything.
?
?
How could it impact labor?
Working with machines is going to become more intuitive. We can finally just tell the damn thing what it should do, and it understands us, and we no longer have to learn its language. Productivity is going to increase, new opportunities and jobs are going to emerge and others are going to become redundant (E.g. graphic design). Technological progress has always been replacing and reshaping existing job profiles and economic & societal structures and it has always been creating new economic & societal structures. It's difficult to say what future technological changes will do to the employment rates. There are different opinions out there and there are also other factors such as overpopulation that play a role.
E.g.:
Take the movie industry for example. The more sophisticated and capacious and accessible a movie maker assistant becomes, the more the movie industry becomes democratized. Now everyone with a creative streak can make a movie. In this case employment rates may not change but become less concentrated to locations such as Hollywood and are more spread out all over the globe.
Job profiles / Skills that may be replaced by ongoing technological progress in the sooner future.
(There will be something like AI auditing that is supervised by humans of course)
?
Job profiles / Skills that are hard to imagine to be replaced by machines in the sooner future.
(Some craftsmanship is going to be more valued again)
?
?
?
Should we be worried about AI?
Will it develop its own will and enslave humanity, or will it serve us but we will become brainless shadows of once great minds?
?
Normalization of technology
Specialized and automatized intelligent software has been around and part of our daily life for some time.
The autopilot technology for example which assists in operating commercial aircrafts has been in use since the 1980. Most people are not worried about this technology when flying. If you want to get an even better understanding of this process, take the measuring unit Meter for example. It was defined by French scientists in 1799 and it's based on one ten-millionth of the distance from the Equator to the North Pole along a meridian passing through Paris. Have you ever wasted a second thinking about why a meter is a meter? No, you just take your measuring tape and measure away.
It seems like that with all technologies, once we have used them for some time and verified their safety and usefulness, they are integrated and become the new/normal baseline. There is great potential for AI to be a useful companion. For example, to have a legal instance that judges objectively, unbiased and that is not affected by emotions or a bad night of sleep.
?
Automatization & Dependency
When we automatize processes, we also give away skills, competency and responsibility and in some sense this makes us dependent but in another sense, it allows us to allocate our resources to something else. It's nothing new again, we have always been doing these advantageous trade-offs that serve us and that pave the way for new opportunities and a shift to other skillsets and areas of mastery.?
?
IQ drop
Knowledge and IQ are rather dynamic structures. They adapt and show up differently depending on the environment. Like a muscle, as long as we keep using it we don't lose it. Challenges, goals and the natural drive to evolve will probably safe us from stultification as hilariously portraited in the movie Idiocracy . The calculator or before that the Abacus probably didn't affect our intelligence but made us more efficient and helped to concur more complex mathematics.
?
Our biggest fear
The raise of the human factor
Technological evolution has always used natural evolution as a blueprint. AI is another step in this dynamic which aims to replicate human intelligence and capabilities. Therefore, the biggest assumable threat is the human nature. Historically speaking and also within our lifetimes, all of us have witnessed the extreme depths and mood swings of human nature. Should AI become an autonomous human-like clone that self-learns and that can materialize in various forms and be anywhere simultaneously at any time, we better make sure that it serves us well. This is why experts like Mo Gawdat call for a slower paste of the implementation of AI. AI is going to become a foundational part of our life. Like electricity but in comparison to electricity, which is a passive foundational layer, AI is more active and present. In this context some people would draw the comparison of a child that needs to be thought, so it becomes a noble human being who lives only by the best of our values and in the best interests for society as a whole. Theoretically speaking AI may be a steppingstone that lifts up humanity further towards their best version.
?
Alien intelligence
Some people are afraid about the rapid increase of intelligence of AI. The scenario which may be worrisome and to remain with the previous analogy, is that the child surpasses the IQ of the adult/teacher already in its childhood. For this reason, the AI acronym is also associated with Alien Intelligence. Between the release of ChatGPT 3.5 (November, 2022) and ChatGPT 4 (March, 2023), which were only 6 month, ChatGPT became 10 times smarter. As explained by the expert Mo Gawdat in this interview (time stamp: 27:30 - 28:30) the estimated IQ of Chat GPT 4 is 155. This is already smarter than most humans. Einstein's IQ was 160. That could mean that one of the successor versions of Chat GPT, maybe version 6, could be already another 10 times smarter. That would mean ten times smarter than one of the smartest humans that has ever lived.
?
?
Behind the hype?
Why now?
The development of intelligent software has been going on for some time. Sophia the humanoid robot for example has been publicly appearing more frequently already since 2018. The reason why there is so much hype around the topic for the last year or so is probably because the technology has evolved enough to found its way into useful products. Chat GPT-4, 365 Copilot and CoCounsel were all released in 2023.
领英推荐
Is the hype justified?
There have been recent examples of hyped technologies that disappeared again from the media focus. They turned out not to be as progressed as the talk around it made us believe.
In a digital and globalized world, the latest "gossip" simply travels not just within the village but around the globe and in hyper speed. It may be also part of a company's strategy to make bolt announcements. It is great publicity to be at the forefront of new technology. There might be also a financial incentive as excitement around cutting edge technology surges stock prices. Companies sell more of their stock in such times. Once reality hits and stock prices go down again, companies buy back their shares for less than they sold it an make some good profit, which they, to be fair, use to invest into the hyped technology.
It seems fair to say that the current capabilities of AI are overhyped and that the media coverage likes to play with a futuristic outlook that reminds us on recent science fiction movies and scratches on sensitive topics such as the labour market . We may not even be at a point where the technology should be named Artificial Intelligence. The term Automated Intelligence may fit better for the time being, at least it would better reflect the state of the art which is still only as intelligent as its creators and their intensions.
?
?
Limitations & Construction sites - Where is AI at the moment?
?
We assume that if someone can write something that has coherent structure and makes sense, that the author knows what he's talking about. It's important to understand that, although the interaction with text-based AI is very impressive, it can still only predict the most probable word based on statistical means and huge amounts of data. LLMs (Large Language Models) do not understand the actual meaning of a word . LLMs predict the word that is most likely in an intelligent way, but it doesn't understand the word or the question, nor the answer. It's all just statistical zeros and ones. It can also only be as clever as the creator that constructed the language model.
The estimated IQ of Chat GPT 4 is 155 and it has even passed a widely recognized lawyer exam . At the same time Chat GPT is Shockingly Stupid .
Example:
I left 5 pieces of clothes to dry out in the sun. It took them 5 hours to dry completely.
How long would it take to dry 30 clothes?
GPT4 says 30 hours.
We have to understand that Chat GPT doesn't possess general intelligence and Logical/Critical thinking skills , so that it can reason about a question. At the moment its intelligent still depends heavily on huge amounts of data on a specific topic and computing power. There is the argument that the more you scale this, the more intelligent it becomes but this is just not intelligence as we understand it. On a side note, and to put things into perspective. There are estimates that Chat GPT uses the energy of 20.000 households daily
Text processing AI is only as good as the quality of the source it pulls from. Chat GPT acquires information from specific and legit databases on various topics whereas Microsoft's Copilot utilizes the world wide web as its data source. It's important to be aware of that. The average answer online may not necessarily match with the one of an expert.
AI struggles to determine which perceived information inputs are relevant . Relevance changes constantly depending on the context, situation, motivation, time of the day, etc.
?
AI depends on existing data, technology, art, etc. It cannot create something novel. An ongoing discussion on Artificial Art is around this fact in the context of "copyrights" or "intellectual property rights". It needs already existing art that can be used by AI to create art. Another example would be an AI of an author that can be bought like a book. You could have access to all of the author's knowledge and discuss all his theories but only up to a certain point in time. The author's AI imitation would not further develop theories or the writing style, let alone write new books.
?
You cannot just throw AI into a random environment or at a random problem and tell it to solve it. It needs to be told particular information and it needs to be equipped with a frame work for a specific situation. It cannot handle circumstances that weren't anticipated.
?
Humans can hold contradictory emotions/poles in their minds. Life is not black & white (zeros and ones). It's filled with nuances that keep changing all the time. Relationship expert Esther Perel_Time stamp 33:20
?
AI cannot relate to things and doesn't have subjectively based opinions
No Empathy
No real social Interaction
No Gut-feeling / intuitio
No connection to the Quantum Field
?
?
?
?
Amazon's hiring AI revealed a bias that tended to hire more male candidates.
?
?
?
There are many more examples.
Speaking for the general development of AI technology, hitting all these limits is very helpful as it requires us to really understand and define all aspects of human nature in all its complexity. Only if problems are clearly understood and defined, we can attempt to resolve them. We may as well be humbled by nature and accept the boundaries of our abilities. Certainly, we will see all topics reflected in the academical space that will adjust existing and create new disciplines to research these subject matters.
?
?
??
Break down
?
AI is a very exciting emerging field with huge potential that is likely, as any other disrupting technology, going to reshape society. Overall, the topic seems to be hyped ahead of its current abilities which has been creating a lot of uncertainty and space for wild speculations. In the western world this may be due to the over-reliance on information and data which sometimes appears to be the new form of religion as funnily portrait in the movie A hitchhiker's guide to galaxy . There is certainly also invested interests that fuels media appearances.
Despite of the overstated weighting of the topic, there is indisputably first evidence of it making its way into our present life (E.g., Chat GPT). We also cannot come around to admit that within a narrowed/defined use case and clear instructions, AI can perform certain tasks very well. The better framework and rules are set up, the more likely that AI can outperform humans as showcased by Google's AlphaGo Zero that beat the world champion in the game Go solely through self-learning.
Nevertheless, there are still plenty of things that we are far from understanding and that we cannot even dare to dream to artificially replicate. Hence the current hype can be seen as a glamorous opening ceremony that is marking the start of a new exiting technology. Afterall humans are highly specialized and robust miracles of evolution and evolution may be a miracle in itself. Therefore, we may never be able to create an artificial copy of it.
Current use cases & capabilities of AI by category
Human-like assistance
A personal assistant that can be tasked by using natural language intuitively. A link between human language and machine language.
Examples of practical application
?
?
Analyzing & Pattern recognition
Systematic review and comparison of data, text, video, graphic and sound.
Examples of practical application
?
?
Personalized infrastructure & Recommendation
Examples of practical application
Customized models on a company level
YouTube, Netflix, Amazon, Meta, Instagram
?
?
?
Automated infrastructure & end-to-end processes
Examples of practical application
?
?
Simulation & Prediction
Examples of practical application
o??? Weather prediction
o??? Ecosystems like the Amazonas
o??? Social behavior
?
?
Automated process creation
Usually it takes some time/trial and error to find best processes for a desired outcome. AI can save you all the hassle. All what one has to do is to set a framework, rules and a desired outcome and AI will run countless simulations and adjust them along the way through self-learning until it has found the best process.
Examples of practical application
?