Just read an amazing insight on LinkedIn from Tomas Moška, co-founder of Turing Society, on open AI topic, interesting read to share with Qoorio community 😊
"AI model once labeled ‘too dangerous to release to the public’ now is 116x larger and soon will be, well ... available to the public.
In Feb 2019, OpenAI labeled GPT-2, an AI language model, as ‘too dangerous to release’. It shocked the world with its capabilities to answer questions on the fly on almost any topic.
Now, they are releasing a superior model, GPT-3. If you give it a task and provide no examples (like translating a text), it has a 76% chance of correctly doing so. If together with a task description, you provided a couple of examples, the chances increase to somewhere around 90%.
The team at OpenAI tested this model by generating entire articles of 200 words, mixing them with articles written by humans, and asking people to mark every article as either AI-generated or real. Results were staggering with only 52% of people guessing correctly. That's a mere 50/50 chance.
Other applications of this model are more creative. Some have already created applications that take an app's description as a task and have GPT-3 generate code (https://lnkd.in/eYAtWcp) and design (https://lnkd.in/eQXncjy) "
So, what do you think? And what would you create with GPT-3? 😉
You can't say it's hard unless you're very good at something. Being conscient of the amount of lessons you will have to have does not mean you are conscient of the real difficulty, and sometimes it's just about starting and seeing if it's really that hard !