Struggling to know what Artificial Intelligence (AI) is and what all these buzzwords mean?
Over the many years that Artificial Intelligence (AI) has been developed and worked on there has been plenty of definitions. Back in 2004 John McCarthy’s (one of the founding fathers of AI) released a Paper that offers a good definition for this new type of technology.
” It is the science and engineering of making intelligent machines, especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable.”
WHAT IS ARTIFICIAL INTELLIGENCE? – John McCarthy
Let’s roll back a few decades before the this definition came into play. The birth of artificial intelligence is linked to the work of Alan Turing. He published a paper in 1950 under the name “Computing Machinery and Intelligence” which basically asks that all important question “Can machines think?” This is where the creation of the “Turing Test” comes into play.
The Turing Test is a test for intelligence in a computer, requiring that a human being should be unable to distinguish the machine from another human being by using the replies to questions put to both. Although over the past few years this test has received a lot of scrutiny it still stands as a good reference point to aim for. No computer as of yet has passed the Turning Test. But who knows in the next few years the way AI is advancing it might start showing Signs of AGI.
Now let’s face forward to today. There are 100’s if not 1000’s of people writing documentation and creating new tests and simulations for AI. Stuart Russell and Peter Norvig publish, Artificial Intelligence: A Modern Approach which is regularly updated and is becoming one of the leading textbooks in the study of AI. So if you are interested in AI it is well worth reading up on it.
AI in it’s simplest form is a field, which combines computer science and robust datasets, to enable problem-solving. You will see words like Machine Learning and Deep learning thrown around within this space. These two disciplines use AI algorithms to create advanced systems that can make predictions and classification based on user input or input stored in their data banks. The data collected has of course caused concerns with a lot of people within the community.
Artificial intelligence has gone through many cycles of hype in it’s life cycle and there always seems something bigger and better just around the corner. But no one saw OpenAI’s ChatGPT coming and it seems to have made some of the biggest sceptics into believers of this technology.
The last time Generative AI was hyped this much was with the breakthrough of computer vision. Where computers could actually view and interact with objects in the real world. Now the focus has been pushed towards natural language processing through the use of Large Language Models (LLM). These new type of generative models are evolving and capable of more than just grammer, they can create computer programs, natural images, generative AI art and provide users with a wide amount of data.
This technology is growing at such speed every day that we are only just starting to explore the potential possibilities it can achieve. There are 1000’s of AI tools already on the market. Of course a lot of them use OpenAI’s API but they can still be classed as AI tools as they are using the API creatively.
But a word of caution, as the hype for AI continues to grow then more business will want to enter this space. Not all businesses will want to put the consumer first so this is why so many people are pushing for regulation around artificial intelligence to ensure the safety for people using these tools.
Not All AI is Created Equal
Just because something is labelled as having artificial intelligence does not mean is using a superior form of it. Majority of AI falls into two different categories. Weak AI and Strong AI.
Weak AI
Look it doesn’t necessarily mean it’s weak as in useless it just means a more simple form of artificial intelligence. Weak AI is AI that is trained and focused around a specific task. Majority of AI that surrounds in our day to day lives is performed by weak AI. Think of Autonomous vehicles, Alexa, Siri, Cortana and such. They perform simple tasks. This type of AI is perfect for the manufacturing industry and labours jobs.
Strong AI
It’s debatable as if we have reached Strong AI just yet as strong AI refers to the use of Artificial General Intelligence (AGI) and Artificial Super Intelligence (ASI). Which as of yet we have no valid evidence it exists. If AGI did exist it would be able to perform any task as if it was human, you would be easily fooled to think this form of AI was a human. As for ASI we are a long way from that, think of Skynet on speed.
AI should always be treated as a tool to improve ones life and daily tasks, not to completely take over our lives. This is a key point we should always remember when moving forward.
Buzzwords Machine Learning and Deep Learning
If you are at all reading about Artificial intelligence or have seen people talk about it you might be all too familiar with the words “Machine Learning” and “Deep Learning”. But you might be asking yourself “what does this all mean?”. Well both Deep Learning and Machine Learning of sub-field of artificial intelligence. Surprisingly Deep Learning is a sub-field of machine learning so it’s ok if you feel a litle lost, it’s all a little bit complex really.
Machine Learning
Machine learning is a sub field of artificial intelligence (AI) and computer science which focuses on the use of data and algorithms to imitate the way that humans learn, the aim is to gradually improve it’s accuracy.
Machine learning is more dependant on human interactivity and monitoring overtime. This is why Deep learning exists to try to aim to automate more of the process.
Deep Learning
Deep learning is a type of machine learning based on artificial intelligence neural networks in which multiple layers of processing are used to extract progressively higher level features from data. Basically a more in depth through version of machine learning. Deep learning is able to grab unstructured data from text or images and then convert that into structured data.
Use of Deep learning allows us to scale machine learning in more interesting ways.
So What Separates the Two
Well when it comes to the differences between machine learning and deep learning we can try to simplify this for you.
Machine learning is the father of Deep Learning, due to it’s older age it needs to be taken care of and monitored. Machine learning is stubborn and is happy doing what it’s used to.
Deep learning the child of machine learning is obsessed with new data and expanding on the data it has. Basically deep learning is “scalable machine leaning”
It really depends on the task at hand. AI tools like Stable Diffusion, Midjourney and of course Bing Chat use a mixture of both. When it comes to generative art though this is where Deep learning really does help, is it allows for a more experimental image.
The Rise of The Machine Generative AI Models
Generative AI are deep-learning models that can take raw data, we are talking about a lot of a raw data say the whole of reddit and “learn” to generate something similar. AI Generative Models work on a set of prompts and produce generated content at a high level. So if you were to train a model on say “picasso” and asked it to create you a picture in the art style of Picasso. It should in theroy create an image similar to his style. If you would like to try this out yourself you can check out our Stable Diffusion Cheat Sheet and start creating your own Generative art images.
What may come as a surprise to you is that for many years Generative models have been used to analyse data and statistics. Yes you heard us, they have been working in the background and you were not aware of it. It wasn’t until deep learning was introduced did data scientists start playing around with the possibilities to generate images, speech and more complex data types.
Chat GPT is leading the race when it comes to generative models and you can already see how successful it has become in such a short amount of time. It’s not the best at generative AI art with it’s Dall-E 2 model that’s reserved for Midjourney. The future for these models is for them to be trained a broad set of data that be used for a wide range of tasks. The aim is to have little fine tuning as we want deep learning to do all the hard work for us. Of course there is still going to be some human involvement to ensure this new technology does not go rouge.
As with all technology that has a use case it is just a matter of time before it becomes affordable to the masses. This is when we will see generative AI enmeshed in a large potion of our daily lives in a bid to help and aid us. As we have said before we need to ensure it remains that way and does not take over our daily lives. As prices drop more businesses will jump on board so expect even your local store to maybe one day be using this type of technology
Some of the uses of Generative AI models can be found below:
- Speech recognition: You may have already used this technology before. You have Speech to Text which is as old as the hills, where you speak into a microphone and your words are saved in a text file. You also have Alexa, Siri and any other assistant piece of technology. If speech recognition didn’t exist you would not be able to communicate with these devices.
- Customer service: Online virtual assistants. A lot of websites are incorporating chat bots into there customer service to handle low level requests. You may have already spoken to one before when visiting a website online.
- Computer vision: If you want AI to interact visually with the world around us it needs to have what is known as computer vision. Think of radiology imaging in healthcare, and self-driving cars within the automotive industry.
- Recommendation engines: Search engines have been using these for years. But as it has improved over the years now even marketing companies are looking to benefit from this technology.
A Brief History Of Artificial Intelligence
The Antikythera mechanism is an Ancient Greek hand-powered orrery, it’s thought to be the oldest example of an analogue computer. It was used to predict astronomical positions and eclipses decades in advance. So this idea of a “machine that thinks” is a lot older than we all think. Ok let’s not delve so back in time, here are some important events that have happened in the evolution of artificial intelligence:
- 1950: Alan Turing publishes Computing Machinery and Intelligence
- 1956: John McCarthy coins the term ‘artificial intelligence’
- 1967: Frank Rosenblatt builds the Mark 1 Perceptron, the first computer based on a neural network
- 1980s: Neural networks become widely used in AI applications.
- 1997: Deep Blue beats then world chess champion Garry Kasparov, in a chess match (and rematch).
- 2011: Watson beats champions Ken Jennings and Brad Rutter at Jeopardy!
- 2015: Baidu’s Minwa supercomputer useS deep neural network called a convolutional neural network to identify and categorize images with a much higher rate of accuracy than the average human.
- 2016: DeepMind’s AlphaGo program, powered by a deep neural network, beats Lee Sodol, the world champion Go player, in a five-game match.
- 2023: A rise in large language models, or LLMs, such as ChatGPT.
Where do you think AI could be in the next 10 years? Why not leave a comment below.
31 thoughts on “What is Artificial Intelligence”