5 min read
GPT (or General pre-trained transformer) 4 is the soon to be released new generation of GPT-3. The software is developed by Open AI, the humanity focused R&D AI company founded, amongst others, by Elon Musk
GPT-3 made headlines by revolutionising the way in which people thought language models could be used and the accuracy that they could deliver.
The new version could be even more revolutionary by the sounds of it.
If parameters are an indication of the complexity and power of this type of deep learning based AI algorithm, GPT-4 will have 100 Trillion of them…. making it 500x the size of its predecessor.
To put in another way, GPT-4 will have as many parameters as the human brain has synapses.
The main output is generating human like text, that could be articles, news dialog and much more.
It terms of much more:
It can write code, whist understanding it. this can allow it to learn and generate systems on its own.
It can be trained to design, by focusing its output on visual outputs to take core elements and either copy or generate new versions of design layouts, websites, mobile applications.
The input for all these outputs is text, for example the user could “ask” the system to make a new website like X but for Y purpose. This can allow very fast understanding of briefs that the system is needed to output without lots of coding or training.
Australian philosopher David Chalmers described GPT-3 as “one of the most interesting and important AI systems ever produced“
The main new improvements of this algorithm will be as follows:
Managing data through quick commands – a company operator or director could type or speak commands to a system controlled by GPT 4 with commands in plain language that could be interpreted by the system to automatically organise, administer or control any part of a company where data or documents are involved. This could remove the need to recruit tech staff to manage these daily tasks.
Design new websites and marketing materials without humans – A trained system could instantly generate a new website layout using pointers to other pages or websites as importation, the GPT AI would then intelligently research and design the best option
Analyse data sets without complex BI system or data visualisation platforms – Instead of spending countless hours installing complex systems a GPT that is linked to a data warehouse could intelligence pull and display analytics for a decision maker to instantly view.
Run customer support channels with next level accuracy – Next level chat bots could deal with customers without having multiple manual intervention needed to support. The system could automatically recruit systems to aid the customer, if given the permission.
Work as virtual assistants with multiple processes and systems – If linked to internal communication systems the GPT could best send messages and documents to the best personnel, with only having typed the request into a port for the GPT.
With leaps of technology such as GPT 3 and the oncoming GPT 4, the next levels of AI benefits are around the corner. Proper installation or utilisation of these algorithms is essential. Many companies have requirements for these systems but don’t realise it yet.
There needs to be a link between the technical applications and the business requirements to facilitate great efficacy based systems for all industries.