| dc.description.abstract | The number of language models is increasing at an exponential rate. Given the rate of improvement for big language models like OpenAI's GPT-3, model size and performance double every 3.4 months, or 100 times every 2 years. When compared to Moore's law, language models are presently expanding 7x faster than CPUs have been for most of computing history!
Rober Dale describes in his paper1 that despite the contrasting opinions about GPT3, it is more appropriate for some use cases than others. Applications that don't include any health risk or involve manual interventions will have better use cases of GPT3. Also, the use of GPT3 doesn't mark the end of humanity. However, we need to be concerned about responsible use. One of the ways is to give access to the institutions with the right use of APIs. Open-ended Applications of the APIs are more susceptible to issues than constrained applications. | en_US |
