In this February Newsletter we want to give you an helicopter view on the immense potential of generative AI, which seems to have taken the tech world by storm.
We will also discuss the potential of AI to accelerate mainstream adoption of digital assets and show basics of how it works.
Apparently, ChatGPT has been a consistent headline so far in 2023.
The generative AI system has captivated the technology world and has been adopted by consumers on a mainstream basis. It has also been able to demonstrate the untapped market for AI and bring to light how far the industry has progressed.
We believe ChatGPT will improve cryptography adoption and education. Specifically, through its advanced communication features to better interact with users.
ChatGPT: it is easy to use and its responses are similar to human responses, they have a very fast pace. When discussing these advantages in the cryptography sphere.
Specifically, arguing that in digital adoption, the biggest advantage is how it can be used to answer questions about how works technology behind them.
In addition, the post stated the fact that ChatGPT had the advantage of being able to help clarify concepts in an interactive and conversational way. Not only that, users stated that they could continue to ask questions until they felt they fully understood any of the cryptography-related things they were interested in.
AI development should be vital to the education sector for these reasons.
Here a self-description from ChatGPT:
© OpenAI
ChatGPT: what the new AI model is and how it works
ChatGPT is the new model developed by OpenAI and is part of the GPT-3 (Generative Pre-trained Transformer 3) models. These are artificial intelligence models based on unsupervised machine learning.
They work by using a deep learning technique known as transformer, which involves using a neural network to analyse and understand the meaning of a text.
ChatGPT, specifically, is part of the InstructGPT family: models trained through deep learning but then optimized through human reinforcement.
ChatGPT was implemented and made public in November 2022.
Due to its power and ease of use it is becoming popular on the web and many, given also the possibility of accessing it for free, are discovering its potential.
ChatGPT is part of the so-called InstructGPTs, which are an evolution of previous models based solely on unsupervised learning.
Instruct models in fact derive from the original GPT-3, but are optimized through human reinforcement.
Specifically, through a technique called Reinforcement Learning from Human Feedback (RLHF). The difference of ChatGPT from other Instruct models (which are also called GPT-3.5 models, such as text-davinci-003) lies in the way through which the model was trained.
In the fine-tuning of ChatGPT, human supervisors were asked to provide both desired questions and answers. These were used as datasets to optimize the ChatGPT model.
ChatGPT is thus a sister model to InstructGPT, which is trained to follow an instruction in a prompt and provide a detailed response.
© OpenAI
Advantages and disadvantages of the ChatGPT AI model
Arguably, the biggest advantage of ChatGPT is its ease of use, as well as the way it can generate human-like responses extremely quickly and (as of February 2023) for free.
The fact that it can generate text in response to almost any request and on a wide range of topics, thanks to its diverse training, means that millions of users have found ways to use ChatGPT to make their lives easier.
For example, developers have used it to help debug code, creative people have used it to help write suggestions for stories or song lyrics, and students have used it to help them understand difficult concepts.
As long as the request can be formulated as text-based input and output, ChatGPT’s functionality is extremely versatile.
Consequently, the existence of ChatGPT can save time and money.
For example, in business terms, tasks that would otherwise require significant human effort, such as answering customer FAQs, could be handled by ChatGPT.
Artificial intelligence can handle a large volume of customer service requests, thus allowing staff to focus on more complex tasks.
This could improve efficiency and reduce costs.
On the other hand, although ChatGPT is extremely impressive, users should be aware that there are a number of drawbacks.
Yes, it is extremely easy to use and produces compelling answers, but anyone using it should not be too complacent or turn off their critical thinking.
It is important to recognize that ChatGPT has limited understanding and lacks basic common sense. Its responses may seem human, but the bot has no real understanding of the text it is processing.
In fact, the language model is simply producing output based on probability distributions derived from its training and data. This means that it can generate misleading or even incorrect information.
Accuracy problems are also related to the quality and substance of the data on which it is trained.
For example, the training data for ChatGPT had a knowledge limit of 2021, so the training process ended before that date. If, for example, ChatGPT was asked about something that happened from 2022 onward, it would have no way of accessing that information.
What can it do?
With its 175 billion parameters, it’s hard to narrow down what GPT-3 does. The model is, as you would imagine, restricted to language. It can’t produce video, sound or images like its brother Dall-E 2, but instead has an in-depth understanding of the spoken and written word.
This gives it a pretty wide range of abilities, everything from writing poems about sentient farts and cliché rom-coms in alternate universes, through to explaining quantum mechanics in simple terms or writing full-length research papers and articles.
While it can be fun to use OpenAI’s years of research to get an AI to write bad stand-up comedy scripts or answer questions about your favourite celebrities, its power lies in its speed and understanding of complicated matters.
Can ChatGPT or similar take over the world recalling some Hollywood most liked movies, we asked to our CEO Federico Corbari: “definitely not ChatGPT as its too commercial to get to that level, but we are maybe opening a Pandora's box” commented Federico and added, “ ChatGPT it has its limitations and its software can be easily confused if your prompt starts to become too complicated, or even if you just go down a road that becomes a little bit too niche but the rout is traced”.
In fact as of today it’s a very useful tool, we could spend hours researching, understanding and writing an article on quantum mechanics, ChatGPT can produce a well-written alternative in seconds.
Equally, it can’t deal with concepts that are too recent. World events that have occurred in the past year will be met with limited knowledge and the model can produce false or confused information occasionally.
OpenAI is also very aware of the internet and its love of making AI produce dark, harmful or biased content. Like its Dall-E image generator before, ChatGPT will stop you from asking the more inappropriate questions or for help with dangerous requests.
How does it work?
© OpenAI
We will definitely come back on the topic, especially on Superintelligence but for the moment we asked to close the February Newsletter to our new AI friend…and here below its the output...
If you have any questions in the meantime please don't hesitate to reach-out to
Digital Assets technologies are in constant (r)evolution; stay tuned to stay updated