GitHub’s Copilot for Business is now generally available

erek

8=D
2FA
Joined
Dec 19, 2005
Messages
18,309
80+% ambitious

"Dohmke expects that soon, Copilot will be able to generate 80% of a developer’s code. Today, that number is about 46% across programming languages — and 61% for Java. Today, ChatGPT and Bing can obviously generate entire applications (or at least parts of them) with just a simple prompt. That’s enabled by the same models as Copilot, so the obvious question, which I asked Dohmke, is when these capabilities will also come to Copilot."

Final.png


Source: https://techcrunch.com/2023/02/14/githubs-copilot-for-business-is-now-generally-available/
 
What will all those poor journos that learned to code do now?
 
Last edited:
80+% ambitious

"Dohmke expects that soon, Copilot will be able to generate 80% of a developer’s code. Today, that number is about 46% across programming languages — and 61% for Java. Today, ChatGPT and Bing can obviously generate entire applications (or at least parts of them) with just a simple prompt. That’s enabled by the same models as Copilot, so the obvious question, which I asked Dohmke, is when these capabilities will also come to Copilot."

View attachment 549065

Source: https://techcrunch.com/2023/02/14/githubs-copilot-for-business-is-now-generally-available/
People are way overhyping GPT, this seems far more useful. GPT can't build whole applications unless your app is fizzbuzz or some crap. Github is not going to "complete your code" if you don't know what you're looking for, you need to know a decent bit about coding to get it to do something useful.
 
  • Like
Reactions: erek
like this
People are way overhyping GPT, this seems far more useful. It can't build whole applications unless your app is fizzbuzz or some crap. It's not going to "complete your code" if you don't know what you're looking for, you need to know a decent bit about coding to get it to do something useful.
Github copilot is also GPT-3 from Open-AI and both GitHup and OpenAI are "partner-co ownership or fully owned" by Microsoft

GitHub Copilot is powered by the OpenAI Codex, an artificial intelligence model created by OpenAI which is an artificial intelligence research laboratory.[10] The OpenAI Codex is a modified, production version of the Generative Pre-trained Transformer 3 (GPT-3),

Microsoft purchase of GitHub for 7.5 billion could end up being on of the great acquisition for them.
 
Yes, exactly as they did.


A type of leech.
What are you talking about? Some journalist wrote a puff piece about people losing their jobs and you formed an identity around it I guess? Still sucks that the workforce is going to get hollowed out, this affects more people than whatever you made up in your head.
 
Github copilot is also GPT-3 from Open-AI and both GitHup and OpenAI are "partner-co ownership or fully owned" by Microsoft

GitHub Copilot is powered by the OpenAI Codex, an artificial intelligence model created by OpenAI which is an artificial intelligence research laboratory.[10] The OpenAI Codex is a modified, production version of the Generative Pre-trained Transformer 3 (GPT-3),

Microsoft purchase of GitHub for 7.5 billion could end up being on of the great acquisition for them.
I was talking about the implementation of the AI, not the API.
 
What’s your opinion of the implementation of the API? Just curious
ChatGPT or CoPilot? I think CoPilot seems a lot more helpful, but again you need to know what you're doing to use it correctly. It's a great productivity tool. GPT seems pretty error prone and I think it's good to build very simple skeleton classes out of and maybe extract some things from it. The best thing its done for me so far is help me solve some math problems in my code what I asked it about specific lines. I went over in another thread how it mishandled assigning a class level object as if it wasn't a reference location and not a value, so it does make a lot of mistakes. It requires some prompting to get things correct. GPT is cool, but I think it's better suited for creative writing and answering simple questions rather than writing whole applications. Overall, I see this as a productivity booster, it's not something that's going to make software devs obsolete. Any company hyping it that way is just doing marketing.
 
Last edited:
  • Like
Reactions: erek
like this
What are you talking about? Some journalist wrote a puff piece about people losing their jobs and you formed an identity around it I guess? Still sucks that the workforce is going to get hollowed out, this affects more people than whatever you made up in your head.
And I comented on exactly those journalists. Why are you so butthurt about it?
 
And I comented on exactly those journalists. Why are you so butthurt about it?
Lol, I'm not I just think it's weird thing to bring up as your first thought, it's almost completely without context and almost a non-sequitur.
 
Lol, I'm not I just think it's weird thing to bring up as your first thought, it's almost completely without context and almost a non-sequitur.
"Learn to code" was in the cultural zeitgeist for quite some time. Now coding, as well as article and paper writing, are getting increasingly automated, which is quite ironic I find.
 
GPT or CoPilot?
CoPilot is powered by GPT, I am not sure of the nuance being made here.

Do you mean ChatGPT everytime you write GPT ?

GPT used in a copilot coding type of context has many advantages, the context is way more defined and limited, the amount of quality input that is the whole open source (and not open I could imagine but hosted on github and other microsoft server) to learn quite good and the user will tend to know if what GPT give it is good or has error in it, versus asking about an information you do not know and that you would have to rely on gpt.
 
CoPilot is powered by GPT, I am not sure of the nuance being made here.

Do you mean ChatGPT everytime you write GPT ?

GPT used in a copilot coding type of context has many advantages, the context is way more defined and limited, the amount of quality input that is the whole open source (and not open I could imagine but hosted on github and other microsoft server) to learn quite good and the user will tend to know if what GPT give it is good or has error in it, versus asking about an information you do not know and that you would have to rely on gpt.
Yeah sorry I meant ChatGPT not GPT, wish they would've named it something else.
 
Meh, AI may automatically be able to generate some code, but you'll still need competent coders to review it and fix it. The benefits will even be questionable in some cases.

It's sort of like a translator using google translate as a first pass, and then going thorough and fixing all the mistakes and things that sound weird manually.

AI will certainly improve productivity in many industries, but I don't believe in mass unemployment in the future.

Every time we have had improvements in automation it has driven MORE employment needs in the industry rather than less.
Take the industrial revolution as an example. The Luddites were pissed that they were going to lose their jobs and sabotaged factories. In reality what happened was that the immense improvements in productivity, resulted in the products being much cheaper, and cheaper products - in turn - drove much higher demand, resulting in the need for more factory workers than there had ever been a need for before.

Similar things happened in Analytical Chemistry labs. Back in the 80's you had to pack your own chromatography columns, run them gravity fed. A single analysis could take many many days and require a highly qualified Chemist to oversee it. Since then pre-packed columns were made available that operate at higher pressures making analysis faster. Machines have been equipped with auto-samplers that can automatically test many vials unsupervised. The chemist just needs to set up the machine and monitor it remotely using a computer.

Does this mean we killed jobs for chemists? Absolutely not. The opposite happened. The data I've found is albeit somewhat sketchy, but there are an estimated 80,000 chemists working in the U.S. today. The number in 1978 was about 5,300. This growth in the number of working chemists has largely been BECAUSE of increased automation. Increased automation resulted in chemical analysis being much more affordable and able to be used for all sorts of things where it previously was just to expensive, and this has resulted in explosive growth in demand, and thus despite being much more productive than they used to be the market needs way more chemists today than before automation. This automation also made many other industries growth possible. There is a reason we have a huge biopharma industry today providing thousands upon thousands of people with well paying jobs.

The truth is, every time in history there has been an large market improvement in automation, overall it has driven more employment, not killed jobs. It may seem scary at the local level, but zoom out and things improve. We may have lost most horse and buggy manufacturers when the car came around, but the highly automated automotive assembly lines demanded thousands of auto workers in return and overall grew the economy.

Jobs didn't go away, they changed, usually for the better. Change can be scary, as it is uncertain, but it doesn't have to be negative.

It is good to be concerned, so we keep our eye on how things are going and can demand our elected officials step in if anything too harmful goes awry, but I bet tools like this will never be able to completely replace software developers. Instead it will make them more efficient at their jobs, making their services more affordable, and used in even more ways than they are today at lower cost, driving even more demand for software developers.
 
Yeah sorry I meant ChatGPT not GPT, wish they would've named it something else.
It make sense has a name, because that mostly what that product added a chat like interface to something that was used for a while by github co-pilot and other Azure AI products.
 
It make sense has a name, because that mostly what that product added a chat like interface to something that was used for a while by github co-pilot and other Azure AI products.
Yeah I get it, it just makes it more frustrating to talk about
 
It's sort of like a translator using google translate as a first pass, and then going thorough and fixing all the mistakes and things that sound weird manually.

AI will certainly improve productivity in many industries, but I don't believe in mass unemployment in the future.

Every time we have had improvements in automation it has driven MORE employment needs in the industry rather than less.

The translating world being an example, giant amount of the translation became semi-automated by AI, it exploded how much translation is being done instead of reducing the workforce, at least for now.
Accounting has well, when excel tab sheet were done by hands, you asked the accounting team 2-3 question a week, yes now what took a day to a team, take 3s in excel to do to a single person, but company ask way more to the accounting department.

we will probably have way more software going on-made, which could keep the number of AI controller-talker needed high.
 
Back
Top