GitHub Universe 2023 opening keynote- Copilot in the Age of AI
Summary
TLDRThomas Dohmke eröffnet mit einer visionären Rede über die Zukunft der Softwareentwicklung durch GitHub Copilot, eine bahnbrechende KI-gestützte Entwicklungsumgebung. Seit 2020 revolutioniert Copilot die Entwicklerwelt, indem es die Produktivität um 55% steigert und eine nahtlose Integration in den gesamten Entwicklungszyklus bietet. Mit über einer Million Nutzern in 190 Ländern und der Einführung von Copilot X für eine umfassende Lebenszyklusunterstützung treibt GitHub die Innovation voran. Satya Nadella, CEO von Microsoft, betont die transformative Kraft von Copilot für die Softwareentwicklung und dessen Rolle als neue universelle Programmiersprache. Copilot Workspace wird als die nächste Stufe dieser Partnerschaft vorgestellt, die Entwicklern ermöglicht, Ideen durch natürliche Sprache effizient in Code umzusetzen.
Takeaways
- 😮 GitHub hat Copilot entwickelt, ein KI-basiertes Tool, das Entwicklern bei der Programmierung hilft und ihre Produktivität um 55% steigern kann.
- 🤖 Copilot nutzt große Sprachmodelle und maschinelles Lernen, um Programmieraufgaben zu automatisieren und natürliche Sprache in Code zu übersetzen.
- 🌐 Copilot ist jetzt in die gesamte GitHub-Plattform integriert und unterstützt Entwickler bei Pull-Requests, Sicherheitsüberprüfungen, Dokumentation und mehr.
- 📲 Copilot Chat wird auf mobilen Geräten verfügbar sein und Entwicklern helfen, unterwegs auf Fragen zur Programmierung zu antworten.
- 🧩 Mit Copilot Enterprise können Unternehmen das Tool an ihre eigenen Codebases und Konventionen anpassen.
- 🔐 Copilot bringt KI in die Sicherheitsaspekte der Softwareentwicklung und hilft dabei, Schwachstellen und Geheimnisse im Code zu finden.
- 🧑💻 Copilot zielt darauf ab, die natürliche Sprache zur neuen universellen Programmiersprache zu machen und den Zugang zur Softwareentwicklung zu demokratisieren.
- 🚀 GitHub arbeitet an Copilot Workspace, einer Vision, bei der Entwickler ihre Kreativität durch natürliche Sprache in Minuten in Code umsetzen können.
- 🤝 Microsoft und GitHub arbeiten zusammen, um Copilot in Unternehmen einzuführen und die digitale Transformation voranzutreiben.
- ✨ Copilot stellt eine symbiotische Partnerschaft zwischen Mensch und Maschine dar, um die Grenzen der Softwareentwicklung zu erweitern.
Q & A
Was ist GitHub Copilot und wie funktioniert es?
-GitHub Copilot ist ein KI-basierter Programmierbeistand, der Entwicklern dabei hilft, Code schneller und effizienter zu schreiben. Es nutzt große Sprachmodelle, um Code-Vervollständigungen, Erklärungen und Dokumentationen basierend auf dem Kontext und den Anforderungen des Entwicklers zu generieren.
Welche neuen Funktionen für Copilot wurden auf der GitHub Universe vorgestellt?
-Es wurden mehrere neue Funktionen angekündigt, darunter Copilot Chat (natürliche Sprachunterstützung in Editoren und mobilen Apps), die Integration von Copilot in GitHub.com (Pull-Requests, Code-Erklärungen), KI-gestützte Code-Scans zur Fehlerbehebung und Geheimnis-Erkennung sowie eine organisationsspezifische Version von Copilot Enterprise.
Wie viele Unternehmen und Organisationen nutzen Copilot laut der Präsentation?
-Laut der Präsentation nutzen über 37.000 Organisationen GitHub Copilot, um ihre Produktivität zu steigern.
Was ist Copilot Enterprise und wie unterscheidet es sich vom regulären Copilot?
-Copilot Enterprise ist eine Version von Copilot, die an die spezifischen Anforderungen und den Codestil einer Organisation angepasst ist. Es wird auf den internen Codebases und Dokumentationen der jeweiligen Organisation trainiert, um ein tieferes Verständnis und präzisere Vorschläge für die Entwickler zu liefern.
Welche Produktivitätssteigerungen wurden durch den Einsatz von Copilot erzielt?
-Laut den Angaben in der Präsentation machen Copilot Entwickler bereits 55% schneller beim Codieren und gibt es die größten Produktivitätssteigerungen im ersten Jahr einer neuen Entwickler-Werkzeugs.
Was ist Copilot Workspace und wie funktioniert es?
-Copilot Workspace ist eine Vision für eine neue Arbeitsumgebung, die es Entwicklern ermöglicht, natürliche Sprache zu nutzen, um GitHub-Issues in Code umzusetzen. Copilot schlägt basierend auf dem Issue einen Lösungsplan vor und implementiert den Code über das gesamte Repository. Es ist für die Zusammenarbeit ausgelegt und automatisiert Teile des Entwicklungsprozesses.
Welche Rolle spielte Microsoft bei der Entwicklung von Copilot?
-Microsoft ist seit der Übernahme von GitHub im Jahr 2018 an der Weiterentwicklung der Plattform beteiligt. Der Microsoft CEO Satya Nadella betonte in der Präsentation die enge Zusammenarbeit zwischen Microsoft und GitHub bei Copilot und der Nutzung von KI zur Unterstützung von Entwicklern.
Wie sollen die neuen Copilot-Funktionen die Softwareentwicklung demokratisieren?
-Durch die Integration von natürlicher Sprache in den Entwicklungsprozess soll der Einstieg in die Programmierung für mehr Menschen erleichtert werden. Laut Satya Nadella könnte Copilot dabei helfen, die Programmierfähigkeiten auf Milliarden von Menschen auszuweiten.
Welche bekannten Unternehmen waren am GitHub Copilot Partner-Programm beteiligt?
-In der Präsentation wurden Red Hat, CACI Corp, Datadog und mehr als frühe Partner für die Erweiterung von Copilot auf andere Tools und Anwendungsfälle genannt.
Wie wird die Preisgestaltung für die neuen Copilot-Funktionen aussehen?
-Alle neuen Funktionen wie Copilot Chat, Integration in GitHub.com und mobile Apps sind im bestehenden Copilot-Abonnement zum gleichen Preis enthalten. Copilot Enterprise für die organisationsspezifische Version wird ab Februar 2024 für 39$ pro Nutzer verfügbar sein.
Outlines
😄 Thomas Dohmke begrüßt die Zuschauer und kündigt eine Disruption der Softwareentwicklung an
Thomas Dohmke, ein Entwickler, begrüßt das Publikum und kündigt an, dass GitHub heute etwas vorstellen wird, das die Softwareentwicklung revolutionieren wird. Er erinnert daran, dass GitHub schon 2020 begonnen hatte, ein automatisiertes Perpetual Gamma-Modell auf Basis von GPT-3 zu entwickeln, das Programmiereraufgaben lösen konnte. Dieses Modell führte zur Entwicklung von GitHub Copilot, das nun von über 1 Million Entwicklern weltweit genutzt wird und zu einer Produktivitätssteigerung von 55% geführt hat.
😎 Live-Demo zur Entwicklung einer App mit GitHub Copilot
In einer Live-Demo zeigt Allison, wie GitHub Copilot dabei hilft, eine neue React-App namens 'My First Commit' zu entwickeln. Sie verwendet Copilot, um den Projektaufbau zu generieren, Code und Tests zu schreiben, das UI zu aktualisieren und Dokumentation zu erstellen. Copilot unterstützt sie durch Code-Vervollständigung, natürliche Sprachverarbeitung und intelligente Vorschläge während des gesamten Entwicklungsprozesses.
🤩 Einführung von Copilot Chat für mobile Geräte und andere IDEs
Copilot Chat, eine Funktion zum Stellen von Programmier-Fragen in natürlicher Sprache, wird nun auf mobilen Geräten über eine native iOS- und Android-App verfügbar sein. Darüber hinaus wird Copilot Chat in andere IDEs wie Visual Studio, Visual Studio Code, WebStorm und andere JetBrains-IDEs integriert. Copilot Chat nutzt nun GPT-4 und bietet kontextbezogene Antworten aus dem aktuellen Editor sowie Slash-Befehle und intelligente Aktionen.
🚀 Einführung von Copilot in GitHub.com und weitere KI-Funktionen
GitHub kündigt die Integration von Copilot in GitHub.com an. Copilot kann nun dabei helfen, Pull-Request-Beschreibungen zu generieren, Code zu erklären, Schwachstellen zu finden und automatisch Sicherheitsprobleme zu beheben. Darüber hinaus wird eine neue KI-gestützte Funktion zur Erkennung von geheimnissen wie Passwörtern und eine Assistent für die Erstellung regulärer Ausdrücke vorgestellt. All diese Funktionen sind in der bestehenden Copilot-Subskription enthalten.
🧠 Demonstration von Copilot zur Programmverständnis und Codeintelligenz
Colin zeigt, wie Copilot dabei helfen kann, komplexe Codebases und Systeme besser zu verstehen. Copilot kann Fragen zu internen Dokumenten, Repositorys und Code beantworten, indem es den Kontext der Organisation nutzt. Durch das Feintuning des Sprachmodells auf die Codebasis des Unternehmens kann Copilot sogar den spezifischen Codingstil und Konventionen der Organisation erlernen. Darüber hinaus kann Copilot über APIs mit externen Tools wie Monitoring, Deployment und Feature Flags integriert werden.
🌐 Partnerschaft zwischen Accenture und GitHub zur Verbreitung von Copilot
Accenture und GitHub arbeiten zusammen, um die Vorteile der assistierten Softwareentwicklung durch Copilot in Unternehmen zu realisieren. Accenture hat Copilot bereits im großen Stil eingesetzt und plant, es für 50.000 Entwickler verfügbar zu machen. Gemeinsam entwickeln sie Lösungen unter Verwendung von Copilot, integrieren es in Accentures Automatisierungsplattformen und verbreiten es in Kundenunternehmen. Dies soll Organisationen dabei helfen, die digitale Transformation durch Steigerung der Entwicklerproduktivität voranzutreiben.
🤝 Satya Nadella von Microsoft diskutiert Copilot und natürliche Sprache
Satya Nadella, CEO von Microsoft, spricht über seine Begeisterung für GitHub Copilot und wie es ihm den Spaß am Programmieren zurückgegeben hat. Er sieht in Copilot ein Werkzeug, das Entwicklern hilft, in einem Flow zu bleiben und Freude an ihrer Arbeit zu finden. Nadella betrachtet natürliche Sprache als universelle Programmiersprache, die den Zugang zur Softwareentwicklung demokratisieren und es ermöglichen wird, Milliarden von Menschen in den Prozess einzubeziehen. Er lobt die Integration von Copilot in GitHub.com und die Zusammenarbeit zwischen Microsoft und GitHub.
👨💻 Die Zukunft der Softwareentwicklung mit natürlicher Sprache
Satya Nadella diskutiert die Zukunft der Softwareentwicklung mit natürlicher Sprache. Er sieht in der Verwendung von natürlicher Sprache einen Weg, die Zugänglichkeit und Zusammenarbeit in der Softwareentwicklung zu verbessern. Nadella ist begeistert von der Möglichkeit, dass verschiedene Rollen wie Produktmanager, Marketing und Vertrieb mit Entwicklern in natürlicher Sprache zusammenarbeiten können. Er sieht in der Integration von Copilot in den gesamten Entwicklungslebenszyklus, von der Ideenfindung bis zum Deployment, ein Game-Changer für digitale Unternehmen.
🔮 Ankündigung von Copilot Workspace für den nächsten Schritt der KI-Entwicklung
GitHub kündigt Copilot Workspace an, ein revolutionäres Konzept, das es Entwicklern ermöglicht, mit natürlicher Sprache ganze Repositorys zu bearbeiten. Ausgehend von GitHub-Issues schlägt Copilot Workspace Lösungen vor, erstellt Implementierungspläne, generiert Code, baut und testet Änderungen. Das Ganze ist für die Zusammenarbeit ausgelegt, mit der Möglichkeit, Vorschläge anzupassen und automatische Fehlerbehebung. Copilot Workspace soll 2024 veröffentlicht werden und einen Schritt in Richtung einer Ära darstellen, in der Entwickler KI als zweites Hirn nutzen können.
🎉 Abschlussworte und Dank an die Mitwirkenden
Zum Abschluss werden die wichtigsten Ankündigungen und Innovationen noch einmal zusammengefasst. Es wird betont, dass GitHub mit Copilot die Softwareentwicklung revolutionieren will. Ein besonderer Dank geht an das gesamte GitHub-Team, das hart daran gearbeitet hat, diesen Meilenstein zu erreichen. Die Präsentatoren Allison, Kedasha und Colin werden für ihre Leistungen gelobt.
Mindmap
Keywords
💡Copilot
💡Künstliche Intelligenz (KI)
💡Natürliche Sprache
💡Produktivitätssteigerung
💡Code-Verständnis
💡End-to-End-Workflow
💡Collaboration
💡Sicherheit
💡Open Source
💡Wettbewerbsvorteil
Highlights
GitHub Copilot has achieved remarkable productivity gains, making developers 55% faster in coding and providing the biggest productivity boost for a new developer tool in its first year.
Copilot is being integrated into every step of the developer workflow, from code generation to pull requests, security scanning, and repositories, turning GitHub into an AI-powered developer platform.
Copilot Chat, with natural language capabilities, is coming to mobile devices, editors like Visual Studio Code and JetBrains IDEs, and the command line, making it a universal programming language.
GitHub is introducing AI-powered code scanning and secret scanning features to automatically fix security vulnerabilities and detect secrets without patterns.
Copilot is now customized for organizations, understanding their internal code, documentation, and repositories, providing personalized assistance tailored to their specific needs.
GitHub is partnering with companies like Accenture, Red Hat, and Datadog to extend Copilot's capabilities and integrate it into their automation platforms and workflows.
GitHub Copilot Enterprise, a new product offering, will bring all Copilot features, including customization and integration, into a single offering for organizations at $39 per month, launching in February 2024.
Microsoft's CEO, Satya Nadella, expressed excitement about Copilot, stating that natural language is becoming the new universal programming language and will democratize access to software development.
Nadella envisions Copilot enabling collaboration between developers, product managers, marketers, and salespeople, all communicating in a multi-agent framework using natural language.
GitHub announced Copilot Workspace, a revolutionary new workflow that simplifies turning ideas into code using natural language, automatically proposing solutions, building plans, and implementing changes across repositories.
Copilot Workspace is designed for collaboration, allowing editing of suggested changes, automatic error catching and repair, and streamlined pull request creation and code review.
Copilot Workspace is envisioned as a step towards an age where developers can use AI as a second brain to stream creativity and turn ideas into reality using natural language.
GitHub aims to bring the intersection of human and artificial intelligence to define the future generations of the AI-powered developer platform.
Copilot's integration into every step of the developer workflow, from idea to deployment, and its customization for organizations, is a game-changer in software development.
The introduction of AI-powered security features, such as code scanning auto-fix and secret scanning, enhances the security and compliance aspects of software development.
Transcripts
Announcer: People of Universe,
Please welcome Thomas Dohmke.
Thomas: Good morning.
Thank you for being here, everyone.
It's so great to see all of you.
I'm Thomas and I'm a developer.
And today is the day
I've been waiting for
for the last eight months.
Today, what we will do on the stage.
How we were build.
What we will launch that will redefine
the essence of software
development itself.
This is a moment long in the making.
Now, for recall,
it was just a few weeks
after GitHub
universe last year
when I had its big bang moment catch up.
It exploded onto the scene.
It washed over the airwaves.
It was the most consequential disruption
to technology
since the advent of the Internet browser.
Our global consciousness
has not been the same since.
But for GitHub, our generative A.I.
story did not start there.
It dates back all the way to 2020,
and it's documented in a white paper.
The internally call the Coding
Oracle Paper.
Back in 2020,
a group of all
wildly talented engineers
got their hands on all version of open
this GPT three model and with this model
they sought to answer a single question
was an automated perpetual grammar fact
or fiction.
Could this transformer model
this strange thought for machine actually
help solve
the tasks of an everyday developer?
Our team knuckle down
and put together a bunch of tasks
in all extended exercises
developers often do in job interviews.
I remember hopping on a call
and one by one
we fed these exercises into the model
and observe what happened.
And in 2020,
out of 203,
handcrafted programing exercises
then prompted the models of 93% of them.
Three years ago, it already solved 93%.
This was the genesis.
This was the moment.
This is what won me over.
And I tell you,
it takes a lot to win over
a skeptical German developer.
And our engineers put
that feeling down in writing
that I've heard and I quote, Change
developer tools in fundamental ways.
In particular,
an automated perpetual gamma can be built
that puts the collective knowledge
of the entire community
at the fingertips of every individual.
Big statement, great early statement.
But we took a risk
and we built the world's
first at scale a purple gamma.
Another tool with a large language model.
Before the world was ready.
And today,
with more
than 1 million page
views of the course, 190 countries
get a copy this now
the most widely adopted A.I.
developer tool in history.
And from this broad
based adoption,
we have seen the most stunning evidence
of productivity gains
since we got rid of punch cards
and assembly language.
Already, corporate is making developers
55% faster and coding it 55%.
Productivity gain,
the biggest ever experience
in the first year
of a novel developer tool.
And people truly love using it.
Open source maintainer,
students, teachers,
developers and Fortune 500 companies
from Shopify to MercadoLibre to Mercedes
to VW
will see in just a short period of time.
GitHub.
Corporate has become
the greatest competitive advantage
in any software driven business.
And now over 37,000 organizations
are using copilot to achieve more.
It's been adopted by organizations
in all sizes
and across all sectors and industries.
Today, the full gamut
between two of them, according
Argo paper, has accelerated
beyond even our wildest aspirations.
But air support it could.
Completion was always
just the starting point.
The change agent for the bigger picture,
you know, many professional developers
actually only code like 2 hours a day.
So what use is it if developers
put blazingly fast in the editor
just to drown in mundane tasks
when the initial coding is complete?
This was the conviction
behind our corporate X
Vision announcement in March.
It was the first view of A.I.
sitting beyond the editor
and infusing
the entire developer lifecycle.
And we did not let that vision escape us.
And in the last eight months,
we created something even bigger
in a world where developer
shortage is on the rise.
In a world
where billions of lines of code
form the foundation
of the global economy,
in a world where developers
are still expected
to keep up with COBOL
legacy code dating back to the 1960s
and yet still build for tomorrow,
we simply have to make it easier
for developers to do it all
and to do it all with a copilot,
just as GitHub was founded on Git.
Today we are re founded on copilot.
A AI is now
the electricity
powering the home
for the world's developers.
In just a short time,
the powerful company
that started it all has evolved.
Expand it
and transform GitHub
into the air power developer platform.
From the idea to the issue to building
and coding
to the viewing,
securing, deploying
to get up the comet self.
GitHub is now
powered by copilot every step
and this is real today
and we're going to show it to you
right now.
Allison: Now, back in March,
we shared copilot shot our vision
for conversational assistance
with all of you for the very first time
built right into your favorite editor.
We designed Shot to help every developer
build the big picture
all in natural language.
We've made shot even more powerful since.
But seeing is believing right?
The contribution graph is cool, but today
I thought it'd be even cooler
to build a new app
called my first commit
com to figure out when mine
and every developer on GitHub
first commit happened.
Now we can see who in here
is the real OG of GitHub.
My least favorite part
about starting a new project
is the starting part.
Who likes a blank space after all?
Thankfully, copilot has me covered
with chat.
I'll be able to explain what I need
all in natural language and skip
all the boring projects set up.
I'm going to open the
chat panel in my editor
to give me a bit more space to read,
and then
we're going to use a new slash command
to generate
an entire project in Visual Studio Code.
If you haven't used slash
commands before,
they're a powerful tool
to help you execute your most common
chat commands with just a single word.
Saving you even more keystrokes.
I've got a list of requirements
already in my head,
so I'll feed those into the new command
and let copilot build the scaffolding
I need.
We're going to use next year
with React components
and our design system primer,
and we'll call the app my first commit.
You'll notice
how copilot gives me
the structure for my new project
so I could create it right here.
But I'm going to jump to a workspace
copilot created for me earlier
that I've already started tinkering with.
Let's see what we're working with here
by heading to simple browser
and typing in Thomas's handle.
Okay, we've got basic page navigation
and most of Thomas's info, right?
Honestly, I think you look better
here, Thomas. Just kidding.
So let's
go ahead and replace this hardcoded image
with the actual picture
from his user data.
I'll select the URL, delete
and copilot comes to the rescue.
I'll save that.
And it's
generally a good idea
to keep tabs on the boss.
So I'm going to add
location here as well.
Also,
copilot is giving me exactly what I need.
It never fails to be magical.
And we also need to add this in the UI
and look
because we added location to our data.
Copilot is smart enough to know that.
Now we want to add it to the page tab.
Save.
Head back to the simple browser.
There's the Thomas we know and love.
Okay.
It's time to grab that first commit now.
It's been a little while
since I've used GitHub search API,
so I'm going to ask
for a little bit of help from copilot.
And don't you love
when you can type perfectly
on stage under pressure?
All right.
We're going to ask copilot.
Give it a second to think.
It always likes to get my heart rate up
just a little bit.
And but it always comes through.
It's giving me exactly what I need.
I'm going to review this really fast.
That looks great.
Head back over here to my editor
and we're going to insert
this with just one click
that'll bug me if I don't fix it.
Now. Great.
We've got that started.
Now I need to include
the commit day here.
Looks awesome.
And once again, we have to update our UI.
All right.
Copilot, 99% there.
But we want our first commit,
so we'll save head back.
Okay, We're almost there.
I'm pretty sure that data is wrong,
and it looks a little funky.
Let's see what's happening.
All right.
Oh, I see what I've done here.
I don't have any ordering
in, so
I'm going to go ahead and select this,
and we're going to
use chat
to help us
figure out how to sort our commits here.
I'll head back and I'm going to say
update to include
to include
or sorry to sort rather
by commit or date.
Give it a second to think.
All right.
That's looking good.
But, you know, when I
when I'm on stage,
I kind of get ascending and descending
mixed up
and I need the oldest commits first.
So I'm going to ask copilot
to revise this answer by simply
asking for the oldest or two newest sort
sought by oldest, newest.
This is great because I can ask
copilot exactly what's in my mind
and it can translate it into the code.
I need.
We'll take a look here.
All right. That looks good.
We'll go ahead and replace the line.
Select save.
Oh, we've got something going on here.
All right, you know what?
I'm under time pressure here,
so I'm just going to replace this with
my trusty demo Save.
If anyone was second
guessing if this was live or not.
You've just got your answer.
All right.
This is looking good.
For 2009.
I was more worried
about getting my driver's
license that year
rather than committing code,
but I really don't need it
down to the minute.
So let's go ahead and clean that up.
And rather than using the side panel
like I've been doing,
I can actually get copilot to help me
write in the editor
right on the line of code
that I'm working on
by pressing command or control.
I so for this line
let's ask to get rid of the time
removed time and time zone.
We'll give it a second to think.
And you know I really love this
need a few
because I can inspect the change
before I actually agree to take it.
And it's in a form
that I'm so used to already
will save head back.
There it is.
It's working exactly as we like.
But that's not all.
We've got one thing left.
My favorite part documentation.
So let's head back.
We're going to select the function
we've been working in
and we're going to bring up
copilot using in line tap one last time
and use our doc slash command.
All right, that looks good.
We're going to.
Except
I got to tell you, I'm exhausting.
I must have pressed
tap like ten times and.
Wow, I need a break now.
Copilot chat is coming
to wherever you are,
whether that's Visual Studio Code.
Visual Studio.
Oh, and now, intelligent web storm.
Hi.
Charm and other jet
brained ideas are available in preview
today.
Chats
in your favorite editors,
but now it's also in your pocket.
We're breaking out of the editor
and bringing Copilot Chat
to your native iOS and Android app.
Pretty cool.
With the power of natural language
on your mobile device,
you can get answers
to any programing question
about your repositories files or docs.
That's Copilot wherever you go,
whether you're waiting in line
at a coffee shop
or procrastinating in the morning
when you can't quite get out of bed.
Copilot chat is now in your I.D.
It's coming to GitHub
mobile
and it's getting way more powerful.
We've upgraded Copilot chat
with a power of GPT four
and even more context from your editor
slash commands and smart actions.
So I'm only ever a click away
from solving my next development task.
And finally in line chat to help you
iterate on coupon code with Copilot
without ever having to move your mouse.
Everything I just rattled off,
including the upgrades to Copilot chat,
chat coming to mobile jet brains
and even Copilot in the seal.
I will be in your existing
Copilot subscription.
Yes,
you definitely heard that right.
If you've got a Copilot subscription
today,
all of these features are coming to you.
But when you might ask next month,
Copilot chat will be generally available
in just a few short weeks.
All of this, taken together,
will make natural language
a new universal programing language
for every developer on the planet.
Kedasha: Up until now,
copilot was just about code generation.
That's changing today.
GitHub is the
AI powered developer platform
from pull requests to security
to security scanning
to your repositories and more.
Copilot is now integrated
into every step
of the developer workflow.
Let me show you what I mean.
So I've been working on scaling the
my first command app
that Allison just started.
It's coming along very nicely.
Take a look.
But I noticed that Thomas has been busy
and he's submitted not one,
but two papers.
So let's take a look at the first one.
Okay, So it looks like there
are a lot of file changes here,
but I'm not seeing any test.
Oh, Thomas.
Coding 55%
faster doesn't mean
you get to write bad code 55% faster.
So I'm going to open up this
in a code space
real quick,
because even though Thomas
wants to get this done quickly,
I don't feel comfortable merging this
without at least one unit test.
And when I open up this in a code space,
I don't even have to touch
my local files.
I can edit and make changes
in an isolated
dev environment in the cloud.
All right.
So I have the user car digest
component open
if I highlight the entire component
pop open copilot chat and type
slash test
copilot is going
to automagically generate
an entire test suite for this component.
Here it is telling me
exactly what it's going to do.
Then it gives me the test
and then it tells me how to run it.
Wow. How cool is that?
Let me save this new test in a file.
I'm going to call it Use a car that
test dogs.
Awesome.
Now let me come at my changes real quick,
because what I'm about to show you
next is a game changer.
Now, what I'm about to show you
for the very first time is GitHub copilot
chat integrated
directly into GitHub icon.
That's right.
Copilot has now expanded
out of the editor. Let's take a look.
So if I go back to the PR
and I go to let's say index.
Yes.
And select lines 25 through 45,
click on the copilot icon
that pops up and type.
Please explain the code.
I always like to say please the air
because you just never know.
But take a look at this
copilot
chat is telling me
exactly what this code is doing.
And it even went
as far as breaking down
the code into chunks.
So it's easy
for me to read and understand
this is the power of copilot
chat in GitHub dot com.
Now let's go back to the PR
because I notice that the description
was a bit lacking.
By the way,
did you know
that around 40% of pull requests
don't have a description?
Yeah, let's fix that.
So we've added a new icon here
for copilot actions.
If I click on the summary command
that pops up, copilot
is going to
help us generate
a detailed PR description.
Copilot is no longer just in your editor.
It is now integrated into PR
is to help you with the mundane stuff
like writing PR descriptions.
Think about how long it takes you
to come up with a meaningful summary
of all your work
as it's given a minute.
Wolff Would you look at that?
Copilot I just generated
an entire description,
very detailed
of all the changes
that we've made so far.
It even included
little clicky links,
but I never would have
bothered to add myself.
Awesome.
Now let's take a look at that second PR.
Okay, so it looks like there's
a security issue
around an unsanitized user input.
We know that.
But GitHub Advanced Security Code
scanning has always been able
to find vulnerabilities in your code.
But now
you don't just get a notification.
There's a complete fix suggested
right here with the power of AI.
This is called code scanning ought to fix
it suggested changes
you need to make
to fix security vulnerabilities.
If I click here,
click commit changes.
And just like that, where secure
this.
This is what we mean
when we say that GitHub is the
AI powered developer platform.
Now I have some great news for you.
This A.I.
powered code
scanning auto Fix feature
is available in preview today,
and we're going even further
with our security offerings.
Are AI powered security offerings
GitHub Advanced
Security has been able to scan
for known secret patterns for a while.
We have more great news for you.
We're introducing a new
AI powered secret scanning tool
available in preview today.
This will help you find secrets
that don't have patterns like passwords
stored in source
control, the cause
of some of the most expensive
exploits over recent years.
Now, as we all know,
the two hardest problems
in computer
science
are naming things
and regular expressions.
So today,
we're making it easier for you
to write your own custom patterns
with our new
AI powered regular expression assistants.
This will all be included
in your company's GitHub
Advanced Security subscription.
We are truly
bringing AI
to every part of the GitHub platform,
and that means every part
GitHub dot
com is the home page
for software development
where every project can begin,
where developers collaborate
and where the open source
community builds and maintains
the world's code.
Now we're integrating copilot directly
into GitHub dot com, so
this means you can dig to code,
generate pull requests
and make the rest of your life
so much easier.
All with copilot.
Colin: As you've just seen,
code generation is an incredible tool.
But one of the most challenging tasks
when building software
is understanding the complex systems
we all work on.
Developers today
work on large,
multi-service, multi-platform, multi
language and often legacy code bases.
To succeed,
we have to build off
a complex foundation,
understanding all that code
that came before us
with GitHub is powerful code
Intelligence copilot
can help us
get the critical information we need
so we can build more faster.
Let me show you how it works.
Now.
Hold on.
I'm logged in here
to my personal accounts.
Let me jump to my work account.
So I can show you
how copilot works for me
as an employee here at GitHub.
I'll use this
brand new feature on GitHub dot
com where I can switch users
with a single click.
That's better.
All right, let's ask copilot a question
about shortening URLs.
How would I create a short link
for my first commit
large
language models
like the ones the power copilot
have strong generalist
knowledge of programing
languages, open source
libraries, and popular frameworks.
You can see here it's suggested
we use BIT.LY.
Now we've heard from organizations
and developers
that you want copilot
to know more about you, your code,
your documentation, your pull requests,
and we've built just that.
Now, Copilot is an expert on
your company's code base too.
Today, what I'm about to
show you
for the very first
time is copilot
customized to your organization
Here, Let's point copilot
at our internal documentation site
at GitHub,
we call our internal Dockside the Hub,
and we have our own internal URL
shortened to ghi0.
I'll ask the same question again.
How would I create a short link
for my first command?
And now copilot is searching
through our internal documentation
and using it to answer a question.
You can see here
it not only knows
the correct chat ops command,
so you able to create a short link
using our internal service.
It's inferring from my question
what data to use in the command.
And it's even giving me an example
of how I can use it.
Let me show you another example,
this time using a project that I work on.
Last year at Universe,
we announced GitHub
all new code search
engine codenamed Blackbird,
and when we built it,
we invented a new data structure
to make search super fast.
Even at GitHub, huge scale.
We call that data structure
the geometric filter.
I'll ask copilot,
what is the geometric filter?
Now,
we haven't
published any information
about it publicly,
but now,
because I've given copilot access
to this repo,
it can answer my questions about my code.
Take a look at the answer.
It's giving us.
Copilot is referencing specific files
and giving us a breakdown
of exactly how this data structure works.
This answer is about as good as you'd get
from any member of the Blackbird team.
Think about what that means with copilot.
You now have an expert available
any time to answer questions
about your code base
right from your browser or ID
And we're not done.
We're taking the next step.
We're now trialing
fine tuning the language model itself
for several large customers,
including AMD.
By taking a model like GPP
for and fine
tuning it on your company's code base,
Copilot quickly learns to adapt
your company's
unique coding style and conventions.
We've seen that fine tuning.
The model for customers
with unique code bases
can bring a massive performance
improvement to their engineering teams
when combined with GitHub
Powerful Code Intelligence
and advanced fine tuning
Copilot will feel right at home
in your organization
and help your company
build even better together.
But what about information
that lives outside of GitHub?
As developers,
we all have a whole ecosystem of tools
like deployment,
infrastructure
monitoring, feature flags and more.
With copilot,
we can now bring information
from all of these surfaces
together in one place.
Imagine right from GitHub
dot com or your idea.
Be able
to check the performance
of a database
query and production
by asking data stacks how it's performing
Using our new APIs for copilot,
developers can extend copilot
for nearly
any use case right within chat.
You can even check the status
of a feature flag using launch darkly
or ask Postman
if you're using an API correctly.
We're kicking off the early phases
of this program
with our debut partners
as part of our new GitHub
Copilot partner program,
including Red
Hat, CACI, Corp, Datadog and more.
We can't wait to see what you'll build.
It's been a year since Chad
Djibouti sparked the general revolution
and kicked off a new era of innovation
in both business
and software engineering. Make one.
Since that moment, Accenture and GitHub
have been working together
to make the promise of assisted
software development real.
At Accenture,
we consider ourselves
Customer zero for GitHub,
we experiment
scale and share our learnings
from our own usage
and client engagements
to help GitHub innovate faster.
We were the first global company
of our size to pilot GitHub copilot,
and we quickly learned a few key things.
First, the promise of efficiency gains
are real
across multiple
phases of the software
delivery lifecycle.
The real value comes from speed
code, quality and security
and developer experience.
And where we see
the biggest opportunity is integrating
copilot into an organization's
end to end automation framework.
Now, Accenture and GitHub are helping
clients move beyond experimentation
to make assisted
technology delivery pervasive.
We are doing this by one core developing
specific solutions using copilot too.
Integrating copilot into Accenture's
own automation platforms.
And three,
by expanding
copilot to 50,000 Accenture developers,
Accenture is proud to be
at the forefront of AI driven innovation
with GitHub
for the
incredible right
50,000 developers that Accenture will fly
at the speed of their own minds
with copilot.
The original copilot
was just code
completions,
wanting inference on every keystroke
with copilot chat
being generally available in December
in Visual Studio and Visual Studio Code
and coming with GPO for inline chat
slash commands
and powerful smart actions.
Copilot expanded from, quote suggestions
to a tool
perpetual camp
powered by one that answers questions,
explains code,
finds bugs and flight tests.
And it's also coming to
your mobile devices
is coming to the comment line.
It's coming to a jet brains ideas
and I'm happy to announce
that all of this
every feature
I just said
is included in your copilot subscription.
And it's the exact same price as before.
Not
$19 per user per month
pocket, is it for organizations?
$10 for individuals.
And of course it's still free
for teachers,
students and maintainers
of popular open source projects.
But on top of all of this,
we've
consistently heard
from many of our enterprise customers
that they would love to have
the internal knowledge
of their organization and copilot.
Today
we are bringing this
customization together
with every single copilot feature
into a new product that we call
GitHub Copilot.
And I place
some copilot
in your pull requests to copilot
in your documentation,
copilot, chat, integrated
and detached to calm all customize
to all of you and your organization
in one single offering.
This personalized copilot
available at every stage
of the software
development lifecycle
By placing the collective knowledge
of your organization's code base
at your developer's fingertips.
They will not only write code 55% faster,
they will keep the flow state
and serve from idea
to code
to deployment by maintaining security,
compliance and quality.
And they will be happy doing it.
And I'm thrilled to share that.
We will start
the preview of copilot in applies today
and it will be discussed
and it will be generally available
in February 2024 for $39.
Copilot Enterprise will give rise
to a new wave of digital transformation
as we know it
and create a competitive advantage
for every organization in our world
that is dependent on software.
But don't just take it for me.
Today we have got a special guest
who knows
quite a lot about digitally transform
being an organization.
Please join me in welcoming
for the very first time ever
get up universe, the CEO of Microsoft,
Satya Nadella.
Oh, hi.
Great to see you, Sacha.
Satya: Thank you so much.
Satya,
I introduced you as the CEO of Microsoft,
but you're really the biggest
super fan of GitHub Copilot.
I am.
And in fact,
I was just
listening to all the announcements
and reading about it.
And I must say in the last,
what, a year, year
and a half, maybe years
since I've been using
Get Up copilot, my weekends have changed.
You know, I can code again.
I mean, the joy of coding is back.
I'd say code space's copilot.
And now I'm really looking forward
to copilot on GitHub dot
com because man, it'll be so fun
to just go to report after report
and keep reading and explain this.
Explain this, explain it all.
Remember to say please. Yeah.
And it
it's been five years, you know
then since Microsoft acquired GitHub
and our industry has changed so much
since then.
How do you all work together?
You know, let us to this moment.
That's cool.
Yeah. It's it's unbelievable.
I'd say five years since obviously
we got to sort of steward GitHub.
You know,
when I think about Microsoft,
Thomas, to me, I'm
always grounding myself, right?
We were a developer company first.
After all, the company got started
creating dev tools.
I always remind myself of that.
And that's our core heritage.
We are a platform company, right?
So we build our platform
so the developers can then build
great IP on top of it.
And we are a partner company,
which is help those developers
be successful in the marketplace
and that's at the core who we are.
And quite frankly,
GitHub has taken us
to another level, right?
Our ability to one, we did a lot of work
before we even decided
that we had the right to steward GitHub.
We made sure that we were great members
of the open source community
by really being part of the community,
contributing to the community,
and that's what led us to that.
So it's been
a fantastic journey
and here we are
at this moment with copilot
and AI and ready for what is,
I think, the next phase
of our developer journey.
Yeah, awesome.
Rene You mentioned a lot of companies
you meet with a lot of developers
around the world.
You get to meet
with a lot of CEOs that hire developers
to drive that transformation.
What do we hear from them
and how can copilot help them?
Yeah, it's fantastic.
I mean, even just listening to the video
you just played
and talked about it,
I mean, think about this.
This is an app that scale organization
like Accenture
making the decision
to go deploy
something like set up
copilot at Enterprise wide.
Oh, that's a big thing.
Usually if you sort of
look at the dev tools business
and dev store
diffusion, it takes decades
before those enterprise
wide things happen.
And this is just happening, you know, in
less than a year perhaps.
A couple of things.
One is at the end of the day, to me,
I think all of us
want to make sure
that the people we work with
have the best tools too,
so that they feel empowered.
They feel that they can stay in flow,
they can do their very best work.
And then,
of course,
the organization benefits
from that in terms of productivity gains.
So unquestionable stuff
is some of the productivity stats.
And the productivity stats, by
the way,
is not just stats, it's about ultimately
some of the drudgery,
bringing the joy back, helping
you stay in flow.
Who doesn't appreciate that, Right?
Every developer appreciates that
and every dev manager appreciates that.
And so to me, that's what we hear.
And really the other thing
that I would say
is I've always felt
like when you introduce GitHub
into an organization,
workflow changes, right?
We always said
like when the salesperson
is, you know, doing a pull request,
it's a different org.
It's moving at the speed of code.
So now think about it, right,
with copilot
and that speed of code
fundamentally changing what it means to
what is a digital economy, right?
If you're in the auto industry,
you're in the financial industry,
you're in the pharma industry
every month.
Pretty much
everybody is a digital company
and a software company.
And so I'm very excited culturally.
How do the organizations change?
How does work and workflow change
across all functions?
It's just not the professional
software developer,
but everyone around them
is collaborating
and really making it
a much more real time
digital organization.
So that's the most exciting thing
that I think
everyone in this room
gets to participate in,
which is
change their organization,
help their organization
achieve next level
of transformation and productivity.
Yeah, incredibly exciting.
And I,
I saw your
I think it was a post on social media
saying that
natural language
is becoming the new universal
programing language.
How do you think natural language
will democratize access
to software development?
Yeah, I mean, look, I think, you know,
you look at all the demos
this morning
and you feel like, wow, I can do that.
I mean, all I need is copilot Chad
and here it is.
I just sort of click, click, click.
And I'm real tap, tap, tap, and I'm done.
I mean, this is like it is in some sense
the most empowering thing
is to be able to give, right?
I mean,
we have, what, 100
plus million
people already on GitHub,
but my dream is like,
how do we empower a billion people
and then the other billions
around these billion people
who are all collaborating
with the power of natural language?
Because if I look
at the arc of computing,
our history of the last
70 years has been
can humans and computers
interact in the most natural way?
What's more powerful
than natural language,
which by the way, starts
with language goes multimodal.
And so there's a lot
to be said about sort of that thing
that you just mentioned,
which is natural language
as the programing language
to me,
and also the other comments
that the people in the morning
sort of said, which I thought struck
a chord with me,
which is to say, wow,
copilot is the new ID.
I mean, think about the approachability,
right?
In Europe,
you have a middle school girl
first learning, coding,
just the approach ability
to be able to say, wow,
you can express some complex
thought and start seeing code
and you're going to get into coding.
That is game changing to me, right?
Because natural language
is about accessibility.
It's about making the barrier
to entry lower
and it's going to be stunning to me.
The other thing that I would also say
is what you're doing,
like some of the stuff
that you're going to show
and demonstrate around,
how about applying it
at the repo scale, Right.
I think you call it workspace.
I think I am so excited about it. Right.
Which is,
hey, it's not just about like
I go from an issue
to code to task to deploy.
If we can really make that entire
flow really magical with natural language
and the product managers,
the marketing folks,
sales folks and developers
are all communicating
in this multi agent framework
that I think
is going to be really next level,
maybe next year we're back here.
We can start talking about
some of that stuff.
Incredible.
Thank you so much for joining us.
Thank you.
Thank you so much.
Okay. Thank you so much.
Wow, What
incredible insights
and what an incredible keynote
it's been today.
We have shown you
how developers can go
above and beyond
for the power of natural language
and copilot.
And we truly think this
AI power
developer platform will change the world.
Thank you
for being part of all of this,
and thank you to everybody at GitHub
that worked
so hard in the last few weeks
to get us to this moment.
And please give it up
for presenters, Alison, Keisha and Colin.
Oh,
but,
but you're right.
Don't don't switch off your online.
I think I forgot something
and I
if you know me,
you know that even in my teenage days,
you know coding back in the 1990s
I've always admired Steve Jobs
and 13 years ago
on this stage here in Yerba Buena, Steve
launched the iPad.
And so I'm so excited to say this.
There is one more thing.
Steve talked a
lot about the concept of machines
accelerating human progress.
He described the home computer
as the bicycle of the mind,
a new partnership
between human and machine
that could take us far
beyond our own inherent abilities.
And today I can't help
but think that we have arrived
at the next phase of this partnership.
Software Developers
Embrace of GitHub
Copilot has shown the world
how humans can symbiotically partner
with intelligent machines
to make big ideas even easier to deliver.
Today we are thrilled to announce
an early glimpse
of a revolutionary new future,
and they're calling it copilot workspace.
Every day, millions developers
start from a very familiar place
GitHub issues.
So we created an air native workflow
that dramatically simplifies
the way you use natural language
to turn these great ideas into code.
We call it the copilot workspace.
Starting with any GitHub issue,
Copilot workspace
automatically proposes
a solution based on its deep
understanding of the code base.
It's your replies, link tasks everything
copilot workspace
then builds a step by step plan
to implement the changes
so that you know
exactly what needs to be done.
And if something isn't quite right,
the spec and plan are fully editable,
so you can easily steer
copilot in the right direction.
Once you're happy with the plan,
simply click, implement
and let copilot
chase down the necessary changes
across your entire repository and beyond.
Simply synthesizing code
copilot workspace builds
and test the changes
and validates that they're successful.
You'll even
run your code,
sit and verify the fix visually.
This looks great.
This workspace is designed
for collaboration.
You can edit any of the suggested changes
and if you accidentally introduce
an error along the way,
copilot will automatically catch it,
repair it and rerun the code.
What's the issues fixed?
It's easy
to create a poor request
with generated summary of your work.
This lets your team review,
merge and deploy faster than ever.
We're building copilot workspace
so you can turn your creativity
into reality every single day.
Copilot Workspace
is just one of the Horizon visions.
Our team at GitHub next is developing
when Workspace is released in 2024,
we'll be consequentially closer to an age
where developers can use
air as a second brain
to stream creativity
in the creation and minutes
of its natural language.
And I think it's just the beginning.
This intersection
of human artificial intelligence,
the continue to define the future
generations of GitHub,
the AI powered developer platform.
Thank you so much for joining us today
and have a great GitHub universe. Oh.
Weitere ähnliche Videos ansehen
Reimagining software development with AI - Shelley McKinley
Besser als ChatGPT? - Microsoft Copilot Pro im Test
So nutzt du Microsoft Loop – ein Überblick der Funktionen und Möglichkeiten
Project, Planner & To-Do – New Features & Previews im Juli 2024
Patch Clamp-Technik / -Methode [Biologie, Neurobiologie, Oberstufe]
CH-47F SR Block II - der zukünftige Schwere Transporthubschrauber der Bundeswehr
5.0 / 5 (0 votes)