Personal Finance
Yesterday
1451
Asking girlfriend of 5months to split bills went BAD
Tech Industry
Yesterday
234
Have you ever come across an actual employee centric company?
Tech Industry
Yesterday
550
Meta Benefits
Tech Industry
Yesterday
1921
How to handle unwanted male attention at work
Tech Industry
Yesterday
838
Why are boomers not retiring
We keep hearing some arguments that chatGPT is not that revolutionary which many people and OpenAI claims it to be. Other than standard “hype of every new thing” argument. Can someone discuss why and how its not that great an innovation and its impact on Future AI. TC - 150k Yoe - 10
It feels like a well designed and developed application of ML models and some great NLP. It's not an innovative technology advancement like others we've had in the past 10 yrs. Don't get me wrong. It's awesome and I use it every day, but it's just that.
Lol imagine using something every single day and being like “it isn’t innovative” 😂
🤣🤣🤣🤣 right?! OMG! What???
Go on Coursera and take the intro to language modeling course by deeplearning.ai taught by Andrew yang. Go through the history of language modeling and how it became what it is today. Your views on gpt3 will change and normalize
I think everyone agrees it is a great innovation and will have major impacts. The disagreement is about it being a job killer. In its current state it is a very useful tool but not good and reliable enough to replace the typical engineer (or other knowledge worker). If it keeps improving at the rate it’s been improving that might change in some years, but for now the “be afraid for your jobs” hype doesn’t make sense.
It will kill exactly 0 SWE jobs. WITCH labor is cheaper and better than AI
True right now. Don’t make assumptions about five or ten years in the future, however. It’s hard for me to see how it doesn’t eventually happen.
It's a LLM and people need to calm down. It can do things more or less correctly but too rough around the edges because there is not an ounce of intelligence (in the usual sense) in LLMs, real or artificial. Regurgitations with modifications.
Until it can update its own source code, we'll be okay
The disruptive thing about chatGPT is the way it was trained. Models trained using RLHF are outperforming models that have 100x more weights. Now this will become the new trend by disrupting the existing trend that more weights mean better model which has been the trend for some years now. Other than that AGI is still far away and far from being a job killer
Nothing about ChatGPT is particularly novel. RLHF and zero shot task transfer were techniques that were in place. Open AI has done a marvelous job of training the model (it is not straightforward) and making it available to the public via an interactive interface apart from brilliant marketing. It is a demo but the demo is also useful. I have not seen any other research lab making a model available for public use where the cost of inference is so high. It is not a step function improvement but it is the best example of packaging an LLM that's useful for a broad set if scenarios.
You mean, other than it’s limitations, isn’t it great?