What GPT-3 Means for AI
Artificial intelligence has taken the world by storm. You can now walk into any house and expect to see inventions that utilize AI technology such as Alexa, Google and Siri. The latest advance in Artificial Intelligence is GPT-3. GPT-3 stands for the Generative Pre-trained Transformer 3, and was developed by OpenAI. GPT-3 is the largest artificial neural network ever created and is revolutionary because it can create things with a language structure. This includes essays, answering questions, computer code and translating languages.
Due to the fact that GPT-3 has 175 Billion Parameters, GPT-3 is capable of meta learning and can do tasks without training beforehand. For example, when a user puts in a question, the question is quickly analysed and the user is provided with the most likely answer. To put it in simpler terms, when we give GPT-3 input, it uses the internet to generate a probable output. According to Open AI, GPT-3 is designed to be “both simple for anyone to use but also flexible enough to make machine learning teams more productive” (OpenAI). Since it is more simple than other forms of AI, GPT-3 can easily be used by the general public which makes it beneficial to larger group of people. Due to its advanced features, GPT-3 will surely be used to solve many problems. For example, GPT-3s ability to replicate humans speaking mechanism could put people with anxiety at ease when talking to AI.
However, GPT-3 is already being used to create amazing projects. For Example, Algolia is combining GPT-3 with their advanced search technology to give their customers the ability to use Semantic Search. For reference, semantic search is an API that allows its users to find something on a website based on its natural language meaning rather than keywords. This would help users find what they need from a website at a more efficient rate. Algolia specifically uses GPT-3 to answer complex queries and trim down the time it takes to do so. While Algolia is using GPT-3 in a truly creative way, there are even more ways to utilize the language. For example, Quizlet (a platform commonly used by students) is using GPT-3 text generation capabilities to create examples of how a certain word can be used in a sentence. Additionally, Koko, an online mental health intervention is utilizing GPT-3 to help identify users in states of crisis and help them.
While GPT-3 brings so much to the AI world, it also poses dangers to low-skilled programmers. Because GPT-3 can program code by itself, it could potentially reduce the need for programmers. For instance, Sharif Shameem was able to use GPT-3 to generate various layouts from simple requests. Since GPT-3 is an extremely powerful neural network, it can code in many different languages. These languages include Python, CSS and JSX. While this feature could potentially increase productivity and make the lives of programmers easier, this would also eliminate the need for low-skilled programmers therefore putting many people out of a job.
Since its release GPT-3, has been receiving lots of attention, mostly due to the fast learning capabilities of the language. Contrary to other models, GPT-3 does not require training before hand, which is proof that GPT-3 is a huge step forward for AI. While GPT-3 is a massive improvement from its predecessors, there is still room for improvement. In the end, GPT-3 is just another example of how far AI has come and how far it will go.
