GPT4 – Potato potahto, or the technology that could change the world?
Last week, OpenAi announced the release of their most advanced GPT model, GPT4. Its capabilities, as presented by the founding company, are way beyond what its younger version was competent to achieve, and it’s no wonder why the entire media was flooded with an unseen excitement, matched only by the previous launch of ChatGPT.
If you want to find out more about our eCommerce projects, we are here to help.
GPT4 is when AI can start making a difference. GPT-4 is a multimodal large language model; it can respond to both text and images and with up to a trillion parameters, it is one of the largest language models ever created. It has enhanced natural language processing, reasoning, and problem-solving abilities, making it even more useful for a wide range of applications.
Why is this so important? Well, mostly because apart from tasks like content or image generation, which of course, created a huge buzz due to the novelty, AI has the potential to help humanity immensely. It might seem like big words, but if we take into account the fact that human errors played a big part in many of last years’ crises, we could see why reducing these errors might have led to a different outcome. So, here are the areas where AI could help.
eCommerce businesses
Leaving aside topics like customer service, marketing and content, or even smart product recommendations based on clients’ past interactions – which we covered in an older article – AI and the new GPT4 can handle more complex and meaningful tasks.
Not long ago, Dutch retailer Scotch & Soda filed a bankruptcy request for its Dutch operations due to serious cash flow problems that began during the COVID-19 pandemic and have continued amid high inflation and a consumer spending squeeze. It might seem as an effect completely out of the company’s ability to control, but at the end of the day, somebody up the command pyramid made the wrong predictions and decisions that led to this unfortunate outcome.
This is where AI and GPT4 can play a critical part. Over time, it can analyze customer sentiment, provide personalized product recommendations, forecast demand, optimize the supply chain, automate customer service, and predict pricing. By making informed decisions based on data analysis, the business can increase sales, revenue, and customer loyalty, while avoiding overstocking or stockouts.
GPT can assist in optimizing pricing and free up customer service representatives to handle more complex issues, reducing the risk of cash flow problems.
Finance
We are on the brink of a worldwide recession, that’s a fact. Over the last couple of weeks, the media has been buzzing about two hot subjects, notably the troubles that Silicon Valley Bank is facing in the US, and the buyout of Credit Suisse by UBS, at a fraction of what its real value was. The buyout prevented an extremely loud crash in the European financial system.
Could AI help prevent it? The answer is yes. AI algorithms can be used to analyze financial statements, market trends, and economic indicators to detect early warning signs of a potential financial crisis. SVB’s customers were largely start-ups and other tech-centric companies that started becoming needier for cash over the past year. Venture capital funding was drying up; companies were not able to get additional rounds of funding for unprofitable businesses and therefore had to tap their existing funds – often deposited with SVB, which sat in the centre of the tech startup universe. AI could have helped Silicon Valley Bank avoid troubles by providing early warning signs of potential risks and vulnerabilities in its loan portfolio.
Now, we’re not stating that AI can solely sort all the problems, but it can definitely help improve decision-making by providing real-time data analysis and scenario modelling, allowing banks to take proactive measures to avoid potential troubles.
Healthcare
There is a well known fact that prevention is cheaper and more effective than treatment because it can eliminate or reduce the need for expensive medical interventions.
Not long ago, Dr. Larry Norton, the medical director of the Lauder Breast Center at the Memorial Sloan Kettering Cancer Center presented an interesting case where AI compared different scans to find abnormalities, thus detecting signs of breast cancer 4 years before it actually developed. AI helped find these abnormalities which radiologists might have otherwise missed.
The use of AI language models in healthcare is seen as an area of immense potential, as it can be used to assist with patient triage and diagnosis, provide personalized health recommendations, and also help with medical research and drug development.
Education
The subject of personalized education is a hot topic right now and AI can definitely help in that instance. If shifting entire national educational systems towards an AI-driven model might seem something out of ordinary right now, it doesn’t mean that it cannot be applied to smaller environments like universities, or online learning platforms.
AI language models have the potential to revolutionize the way education is delivered and assessed. Teachers can use them to analyze vast amounts of data and provide personalized feedback to students, this way enhancing learning outcomes, improving student engagement, and reducing teacher workload.
AI has the potential to help us perform better. A concrete example is the integration of GPT technology into the field of AI-assisted coding, which can revolutionise the way developers write code, making it better, faster and more efficient.
What is more interesting is the fact that AI is capable of self-learning. It can learn from its own data, its own actions and feedback, to adjust its behavior and improve its performance over time.
And with the release of the latest GPT4 model, this behavioural adjustment will be incredibly fast. Most researchers agree we’ll eventually reach a point where our AI systems begin to exceed human performance at virtually every economically valuable task, have transformative effects on our economies, our societies and even our species.
Law
The legal industry is known to be ridden with various necessary but time-consuming tasks. Done by humans, they are prone to errors. AI language models can be used successfully to review and analyze legal documents, such as contracts and patent applications, to identify key clauses and patterns, or identify patterns in previous rulings. Overall, AI can help legal professionals to make more informed decisions and save time on research and analysis.
Available products
Since 2015, when OpenAI was founded by a group of technology leaders including Elon Musk, Sam Altman, Greg Brockman, Ilya Sutskever, and Wojciech Zaremba, the company worked assiduously for promoting friendly AI for the betterment of humanity.
The partnership with Microsoft, which invested billions in the company, allowed for several breakthroughs, including the creation of an AI-powered supercomputer called GPT-3, which was, at its time, one of the most advanced language processing systems in the world, with 175 billion parameters (configuration variables that are internal to the language model, essential for generating predictions). Currently, no specifications are displayed regarding the parameters used in GPT-4, but there are speculations that it exceeds 1 Trillion.
Even if it is not open-source, GPT3 is widely available and accessible as an API for developers and can be integrated in different kinds of applications. GPT4 was launched and is being rolled out gradually, based on a waiting list.
OpenAI has several products that employ GPT. Some of the most used are Codex, an AI model designed specifically for programming, which can understand natural language inputs and generate code in various programming languages, and DALL-E. The latter can generate images of objects, animals, and even scenes based on written descriptions.
At the same time, as main investors into the program, Microsoft has introduced GPT into its Microsoft365 apps to unlock productivity. Microsoft also expressed plans to integrate OpenAI’s GPT technology into its popular code-writing tool, called Visual Studio Code with Copilot, allowing developers to use natural language to describe the code they want to write, and then the tool generates the code for them. This has the potential to revolutionize the way developers write code, making it faster and more efficient.
Even if the spotlight was stolen by OpenAI and their products, there are some open source alternatives. EleutherAI’s GPT-J is an open-source alternative to OpenAI’s GPT-3. The model is significantly smaller and only contains 6 billion parameters. GPT-NeoX is another open-source language model released by EleutherAI. This model exceeds GPT-J’s size but is still smaller than GPT-3, containing 20 billion parameters.
These are models available to the public. On the other hand, there is Megatron-Turing NLG, created by Microsoft and Nvidia, the largest and the most powerful monolithic transformer English language model, featuring a bidirectional architecture, making it capable of better understanding the context of a sentence than GPT. Megatron Turing NLG is trained on an even larger corpus of text data that includes academic papers, books, and other sources, as opposed to GPT which uses only internet resources. At the moment, it is the most advanced Natural Language Generation model. To accelerate research, Nvidia announced an Early Access program for its managed API service and is inviting organisations to collaborate on research problems.
Google’s response
The tech giant is trying hard to keep up with its rivals in this new AI race. Although the company requested AI tools to be added to all its user products, these came in the form of AI writing tools in Docs and Gmail, which were made available to a group of US-based “trusted testers” first. Right when we were compiling this article, Google released BARD, their response to ChatGPT, but only to users in the UK and United States.
Google’s product had a shaky start after it appeared to give an incorrect response during a demonstration video. This resulted in a one hundred billion dollar drop in the company’s market value, proving that when it comes to AI, Google still has some work to do. But it will be pretty interesting to see how it can match its renowned rival, adding the fact that the latter will benefit from GPT4 which we estimate will be made available to all users by then.
And finally, there’s Apple
While other technology companies have already spent billions on AI, Apple seems strangely reticent. Or maybe they realized the race is already won and there is no point spending resources on developing a fast solution.
For sure, Apple will continue working toward its own solution, but this could take years. And in the meantime, developers have managed to create apps to add ChatGPT to Apple’s products, such as Petey – AI Assistant. Therefore, Apple might simply suck it up and make it easy for its developers to add support for OpenAI’s tech in the apps they make, effectively passing the cost to them and their customers. And why not, spend its resources reinventing, for God knows how many times now, their gadgets as the ultimate Personal Computers. Not as the same machines with apps like Microsoft Copilot added, but as devices which can act as personal assistants which can communicate important information, write, read, hear.
Of course, there is a “but” – as we know it, GPT systems with billions of parameters require massive cloud infrastructures. So, how could Apple make it run on a personal device? Well, it seems things are moving quite fast in this direction too. Georgi Gerganov, a software developer, has developed a program named “llama.cpp” which enables the execution of LLaMA, Meta’s latest GPT-3-like AI language model, on a Mac computer. Later on, individuals figured out a way to run LLaMA on a Windows operating system. Furthermore, an individual demonstrated the model running on a Pixel 6 mobile device. If this keeps up, Apple might have found its solution. Of course, we’re not quite there, but in the fast lane of AI development, things seem to move rapidly.
Another worry is privacy. Nobody would want all its life stored somewhere in the cloud. If Apple could sort this out and find a way to process everything at personal computer level, without compromising privacy, we might witness another breakthrough of the tech giant.
Conclusions
News about AI developments came at the same time as reports about certain financial institutions having troubles surfaced. To some extent, this somehow sweetened the somber predictions about an inevitable recession, because everyone now understands that AI can help us react extremely quickly to the changing environment and make better decisions.
With the new capabilities of GPT-4, AI has the potential to revolutionise the world by enabling unprecedented levels of automation, efficiency, and accuracy in various industries.
GPT-4 will take natural language processing to a new level, allowing for more sophisticated communication and interaction between humans and machines.
With some of the most acute crises happening in a short time span, over the last couple of years only, that’s what humanity needs for sure: better informed predictability in a highly unpredictable world.
If you want to know more about our list of services at Clever++, please visit our website.