ILLUMINATION
Published in

ILLUMINATION

I asked AI about human races. I didn’t like its answer and neither will you.

Image designed by the author

Forget about the GPT-Neo. Start using GPT-3

This was considerably the first sentence I read about it. I remember looking for web development instruction when I came to throw this. The news was sealed by this new algorithm provided by Open-AI and how It has amazingly shaken the AI world!

While I was already involved in web development and ethical hacking for years, my enthusiasm about AI and what happened wouldn’t give me a reasonable reason not to explore it. I decided to glimpse at the subject, but first, I had to explore the AI world and how the concepts were attached — So, here is my journey about AI and GPT-3 algorithm.

We cant solve problems by using the same kind of thinking we used when created them — Albert Einstein

I’m learning effectively — Eh, I guess?

The book 20 first hours by Josh Kaufman declares how our brain has this learning peak phase and how we can learn anything fast within the first 20 hours. The book talks about model sets and how important they are in learning. Model sets are unique phrases used in the profession — pretty much like inside jokes between two people. For Example, in web development, we use terms like server, database, domain, and the like, which each of them have their description. These model sets lead’s us to smaller model sets and help us complete our knowledge tree. This is the method we are going to use during this journey, and as you see, it goes up and never goes downsounds like a 1990’s view of the real state market to me!

Beginner’s Guide to Artificial Intelligence

As a child, AI usually would sound like a robot world to me. Sounds simple, and it is. It’s about how computers process thoughts and put them into action. We can see Artificial Intelligence everywhere — Computer Games, Education, Law, Manufacturing, Business, and on and on.

Hint! Read for Key Details. Big Ideas need support.

Let’s break it down:

image by author
  • Experts Systems. It’s a computer system that emulates the decision-making ability of a human expert.
  • Machine Learning. Using mathematical models of data to help a computer learn without direct instruction.
  • Deep Learning. Imitates the way humans gain certain types of knowledge.
  • Neural Network. Series of algorithms that endeavor to recognize underlying relationships in a data set.
  • Fuzzy Logic. Intended to model logical reasoning with vague or imprecise statements.
  • Natural Language Processing. The application of computational techniques to analyze and synthesize natural language and speech.

Wait! we didn’t include robotics?

Robotics includes a physical form robot and may or may not have Artificial Intelligence. Only a tiny part of robotics involves AI. Most industrial robots are non-intelligent. In Short — It’s a Complete different branch, but it can be linked with AI.

Deep Learning is a branch of machine learning, but we will take care of it since its AI-related.

Expert Systems

Expert systems are involved in a unique structure that can solve complicated issues by reasoning through bodies of knowledge.

The human expert provides knowledge for the knowledge engineer. The knowledge engineer creates an AI system with the available knowledge base from experts to make decisions, then an engine to act on data and a user interface to interact with the user. In this figure, you can see the chain of actions for expert systems.

Here are some examples of Expert Systems:

  • MYCIN: It was based on backward chaining and could identify various bacteria that could cause acute infections.
  • DENDRAL: Expert system used for chemical analysis to predict molecular structure.
  • PXDES: An Example of an Expert System used to indicate the degree and type of lung cancer.

The system automates the process by taking action on human expert knowledge.

Machine Learning

In Machine Learning, we can have a prototype to differentiate between apples and oranges. You have data that says that oranges weights between 150–200g and apples between 100–130g. Also, oranges are rough, and apples are smooth. If you have a fruit that weighs 115g and is soft, your program can determine that it is probably an apple. Vice-versa, if the fruit is 175g and rough, it is most likely an orange. The more data, the more accurate results.

Machine Learning is a form of Artificial Intelligence in which the program is designed to learn on its own.

Deep Learning

There are two ways to do artificial intelligence generally. The first way is to develop each possible situation you want the computer to recognize. Speech Recognition software like Siri and Google Assistant are great examples for this case. For Example, companies have to tell their system what to do in every possible situation for these services.

  • If you hear [THAT], do [THAT]
  • If you hear [WHAT TIME IS IT], open the clock app for the [TARGET] timezone.
  • If you hear [HOW ARE YOU] respond with [IM FINE, HOW ARE YOU]

The other form is called deep learning, which tries to avoid this by only instructing parts of the speech.

Deep Learning learns what to do, but not by following a set of rules.

Neural Network

Neural networks are utilized to identify patterns for AI to take action on. These networks are similar to Neural Nets, simplified models of brains. The Neural net contains neurons that are either activated or inactivated and linked together with the help of synapses.

Neurons are positioned in layers — Input Layer, Hidden Layers, Output Layers. When we are attempting to learn a new matter, we present new data to the input layer, then the activated neurons trigger inactivated neurons to start them, and so on; Until we reach the output layer where the brain network creates a prediction.

If the prediction is correct, it will pass; our neural brain nets rerun the process until we reach the right statement.

Neural Network is based and functions like a brain’s neuron system.

Fuzzy Logic

Fuzzy logic allows a computer to understand things that don’t have a definitive meaning. If we gave a computer our height and wanted to know if we were tall or not, a programmer should tell the computer what measurements are considered Tall.

Without fuzzy logic, the output for whether I am tall or not is only — yes or no. Fuzzy logic gives the computer a way of saying Kinda tall, slightly taller than average, etc.

Natural Language Processing

While growing up, we learn how to speak, we start with simple words, and over time our minds understand more terms and the relationships between them. In many cases, we need computers to understand our natural language and process them for communication between the AI System and the user.

  • Smart assistants
  • Search results
  • Predictive text
  • Language Translation
  • Digital phone calls
  • Data analysis
  • Text analytics

Natural Language Processing creates AI models that process and understand natural language for the AI and regular user communication.

There is nothing so useless as doing efficiently that which should not be done at all — Peter Drucker

GPT-3 Algorithm

We have already covered a lot in AI, and it’s time to move on with GPT-3. GPT-3 belonged to the subfield of natural language processing and was developed and launched by OpenAI. It is based on a massive neural network with 175 million synapses and can add and compose texts under minimal specifications. The largest Natural Language Processing Transformer can accurately and effectively reproduce human thought and reasoning patterns in text. The generated texts are so well composed that readers can no longer distinguish them from texts written by humans. Since GPT-3 was launched in 2020, it has generated an average of 4.5 million words a day and is used by more than 300 apps.

Let’s take a look at these examples:

  • Olive oil is oil made of olives
  • Baby oil is oil made for babies

We want AI to make distinctions like that, but we don’t want to write out all the rules by hand for every possible situation; that would take forever! So instead, we try to make the AI figure that out on its own.

GPT-3’s performance is on par with the best language models for text generation, significantly better than previous GPT models. GPT-3’s architecture consists of two main components: an encoder and a decoder. The encoder inputs the last word in the sentence and produces a vector representation of it, which is then passed through an attention mechanism to make the next word prediction. The decoder inputs both the previous word and its vector representation and outputs a probability distribution over all possible words given those inputs.

GPT-3 is a deep neural network that uses the attention mechanism to predict the next word in a sentence.

All About Experiments

OpenAI shared the beta version because they wanted to see the potential GPT-3 could provide and what new performance we can see. As expected, in no time, Twitter and other blogs were flooding with outstanding results from GPT-3.

Despite its new tricks, GPT-3 is still prone to spewing hateful sexist and racist language. Fine-tuning the model helped limit this kind of output in GPT-2. Unfortunately, the algorithm can’t realize which of its results are right and wrong; it has no way of stopping itself from deploying inappropriate content into the world.

The Ins and Outs

GPT-3 has the potential to automate tasks that require language understanding and technical sophistication. For example, it can — Code, Chat, Write Applications, Create Content, Write an Email or Blog Post and even Play Chess with you. But on the other hand, the main problem is that GPT-3 learned about language from the internet. Therefore, its training datasets do not include — News, Articles, Wikipedia, or Online Books, but also every unsavory discussion on Reddit, Quora, and so forth. Since this is troublemaking, in some examples of generating fake news, lying, and along with others— Ending up with harmful biases.

Putting it into action

Busyness is a form of laziness — Dan Lok

GPT-3 produced terrific outcomes, received fantastic publicity, induced nervousness, and received a wave of commentaries. You may want to bookmark this page so you can come back to it in the future when you’re struggling with this Open-Ai new GPT-3 algorithm. You can also use this path today to brainstorm countless ideas for challenges that could be done for your projects by this great algorithm.

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Ashkan Ebtekari

Ashkan Ebtekari

Creative Developer Located in Persia | Software Engineer and Ethical Hacker