#75 Demystifying Artificial Intelligence by describing GPT-3
AI is a suitcase term.
Lot's of stuff gets packed in there which can make it tough to understand.
In 300 words, I'm going to try and describe GPT-3, a recent and massive AI development.
'In the broadest sense, AI refers to machines that can learn, reason, and act for themselves. They can make their own decisions when faced with new situations, in the same way that humans and animals can.' - MIT Technology Review
1 - GPT-3 is the most powerful language model ever
It was fed most of the internet as data to train itself on to develop a gigantic leap in language generation capability. GPT-3 uses all this info to do a simple task: guess what words are most likely to come next, given an initial prompt / direction.
2 - GPT-3 can create anything that has a language structure
For example, it can answer questions, write essays, summarize long texts, translate languages, take memos, and even create computer code. It's a neural network that outperforms all previous NLP models on a range of challenging problems. In other words, it makes computers that can read and write.
3 - GPT-3 learned via unsupervised learning, how humans learn
Instead of requiring large, labeled data sets and predetermined ideas for what to look for / desired outputs etc (supervised learning), GPT-3 learned without being provided with a clear answer to the problem looking to be solved (unsupervised learning). Just a bunch of unlabeled data aka figured it out. Implications here are significant as there's a whole lot of unstructured data out there...
4 - Application of GPT-3 in creative tech
To make it real, let's focus on Copy.ai, a Craft Ventures backed start-up, that provides automated creativity tools built on top of GPT-3. Think about the amount of time spent generating digital content today - digital ad copy, social media content, website copy, eCommerce copy, blog content, sales copy etc. Then think about #1 - #3 above and how much more / better content you can create faster, how much more / often / better you can iterate, etc. Talk about a productivity gain...
Humans + machines. Not humans vs machines.