There's a huge new AI paper that signals what AI will REALLY do to white-collar work.
Hint: ChatGPT is going to seem like child's play by the end of the year.
It's called “Toolformer: Language Models Can Teach Themselves to Use Tools“. And it does exactly what it sounds like.
What it does: since ChatGPT is bad at math, the Toolformer version would know that if it wants a percentage, it needs to type in “400/1400“ in a calculator application, then spit out the answer: 29%.
To say it again: the model knows (1) that it has to use a tool, (2) which tool to use, and (3) HOW to use that tool. This is HUGE.
Businesses have to be very excited to see this track of research play out. There is a mind-blowing amount of white-collar work that is simply typing various things into different systems:
- Updating project trackers
- Data entry into CRMs
- Compiling reports
So imagine the “business“ version of Toolformer. Imagine typing in “Close the deal with Adventist Health and send a summary to the company Slack“.
And it will find the deal in Salesforce, mark it Closed Won, write a summary, then post it to Slack in the right channel.
When we talk about AI's automation potential, this is what we actually mean.
There's going to be a LOT of money made by whoever can pull this off.
PS. Want more like this? 16K+ professionals read my daily AI digest, where we break down the top stories of the day. Join here: http://bit.ly/3iKiI10
#artificialintelligence #machinelearning #technology
Hint: ChatGPT is going to seem like child's play by the end of the year.
It's called “Toolformer: Language Models Can Teach Themselves to Use Tools“. And it does exactly what it sounds like.
What it does: since ChatGPT is bad at math, the Toolformer version would know that if it wants a percentage, it needs to type in “400/1400“ in a calculator application, then spit out the answer: 29%.
To say it again: the model knows (1) that it has to use a tool, (2) which tool to use, and (3) HOW to use that tool. This is HUGE.
Businesses have to be very excited to see this track of research play out. There is a mind-blowing amount of white-collar work that is simply typing various things into different systems:
- Updating project trackers
- Data entry into CRMs
- Compiling reports
So imagine the “business“ version of Toolformer. Imagine typing in “Close the deal with Adventist Health and send a summary to the company Slack“.
And it will find the deal in Salesforce, mark it Closed Won, write a summary, then post it to Slack in the right channel.
When we talk about AI's automation potential, this is what we actually mean.
There's going to be a LOT of money made by whoever can pull this off.
PS. Want more like this? 16K+ professionals read my daily AI digest, where we break down the top stories of the day. Join here: http://bit.ly/3iKiI10
#artificialintelligence #machinelearning #technology