2024 UPDATE!! Altman: prepare for AI to be "uncomfortable" 33% US jobs gone..SKYNET, AI medical advances? BASIC INCOME? 1st AI MOVIE! AI MAYOR!!

[bnew]

I didnt say they trained it on strictly their own interview question data


I said that AI being able to regurgitate (via a LLM) problem solving of problems that are typically presented on an interview question doesnt create an AI worker capable to create, maintain, update, support and iterate on a software product of even small-size codebase.

right now o1 is a template-creator at best, trained on public data, including repositories. it is not trained on proprietary libraries and it doesnt have reasoning capability (nor does any generative LLM) and the more obscure/undocumented a library, API or code is - the worse the AI learning be. After all, an LLM is ALWAYS only as good as its training data. It cannot reason nor can it deduce nor does it have a state-based knowledge. the whole point of transformers is for them to be stateless.



it can help to create some skeleton based on copycat approach right now for folks who dont know or know basics of programming, but thats about it. it may help you learn a language faster or create very skeleton-y code.


I think every time we talk about LLMs taking real programming jobs, we need to emphasize that LLMs are stateless by their nature. Transformers are stateless. The only way state is passed is via context window. This is a big, gargantuan problem with maintaining and adding and having siloed knowledge that is up-to-date, correct and compartmentalized . Right now there are apis to feedback the prompt history each time, but this isnt a real solution with real programming tasks.


for example, I have had a real-world task in maintaining a codebase across several different legacy repositories such as clearcase, and making fixes to proprietary C libraries that only exists in my company but is used by dozens of different products/components. I also need to know how the library is used by these components, and the ability to intrerpret how this library may be used, how often and what issues can arise depending on what component uses it.


only in fantasy world does someone sit down on a programming job and solve interview tasks. real world doesnt work like that. Not to mention, LLMs rely on human language for prompt. In programming, so many things cannot be easily defined or specified in rigidity of language. many things require visual feedback (web development for example or texture creation and rig wiring in graphics) or can be described succinctly in diagrams but not language.

I dont see real programming jobs with people who have a lot experience being taken AI anytime soon. Maybe down the road when someone attaches state, logic to actual reasoning and not merely inference.

https://openai.com/index/introducing-openai-o1-preview/

September 12, 2024

Introducing OpenAI o1-preview​

A new series of reasoning models for solving hard problems. Available now.

Update on September 17, 2024: Rate limits are now 50 queries per week for o1-preview and 50 queries per day for o1-mini.
We've developed a new series of AI models designed to spend more time thinking before they respond. They can reason through complex tasks and solve harder problems than previous models in science, coding, and math.

Today, we are releasing the first of this series in ChatGPT and our API. This is a preview and we expect regular updates and improvements. Alongside this release, we’re also including evaluations for the next update, currently in development.

How it works​

We trained these models to spend more time thinking through problems before they respond, much like a person would. Through training, they learn to refine their thinking process, try different strategies, and recognize their mistakes.

In our tests, the next model update performs similarly to PhD students on challenging benchmark tasks in physics, chemistry, and biology. We also found that it excels in math and coding. In a qualifying exam for the International Mathematics Olympiad (IMO), GPT-4o correctly solved only 13% of problems, while the reasoning model scored 83%. Their coding abilities were evaluated in contests and reached the 89th percentile in Codeforces competitions. You can read more about this in our technical research post.

As an early model, it doesn't yet have many of the features that make ChatGPT useful, like browsing the web for information and uploading files and images. For many common cases GPT-4o will be more capable in the near term.
But for complex reasoning tasks this is a significant advancement and represents a new level of AI capability. Given this, we are resetting the counter back to 1 and naming this series OpenAI o1.

---

---

i'm not a coder but there have been a few times where i ran into issues of various LLM's not knowing some api calls even when they're well documented, i usually just paste as much as i'm allowed to see if it can make sense of it and sometimes that works for me. other times i just modify the prompt to instruct it on crucial info to jump start a more accurate response.

these AI companies including openAI offer fine-tuning services for their models so companies can train their proprietary libraries and other data when RAG isn't good enough to take on the large amount of data they have.

the real solution to stateless if not fine tuning the model is increasing the context window which has been happening albeit not as fast as people would like.


to your point of visual feedback to be able to properly address certain tasks, text-to-text large language models hav evolved to multi-modality large language models that support text, images, speech and video. I know exactly what you mean though because I've gone don the rabbit hole of trying to describe visual modifications I wanted in desktop apps and webpages and it can be frustrating with natural language alone but not entirely impossible.

I think chatgpt Memory feature is how they're currently augmenting state based knowledge into their model. I actually like how transformers are stateless because the chat can go into a direction i don't want it to or be contaminated with bad responses I can't remove that taints further responses.

i don't think a larger context window is better than a state based model especially when dealing with information that changes constantly.

I agree with you that logic/reasoning will need to greatly improved for these models to make real headway into software engineering and other logic heavy domains.

what do you think of the progress being made on agents & models for solving github issues? SWE-bench

 

[bnew]

 

[GnauzBookOfRhymes]

Dockworkers at ports from Maine to Texas go on strike, a standoff risking new shortages

The strike affecting 36 ports is the first by the union since 1977. The strike will likely have an almost immediate impact on supplies of perishable imports like bananas, for example.

apnews.com

Dockworkers striking over automation.

 

[bnew]

Spoiler: full text

Longshoremen in China:

The Chinese port in Guangzhou.

Virtual Longshoremen unload ships remotely with 5G, and help supervise AI vehicles that automatically drive the containers to trucks and load them, all remotely.



To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

 

[bnew]

 

[Lord Beasley]

@Artificial Intelligence you need to chill crodie

 

[bnew]

Wimbledon: All England club to replace all 300 line judges after 147 years with electronic system next year

Wimbledon will replace all 300 line judges from next year with AI technology; Watch all the ATP and WTA action on Sky Sports Tennis and Sky Sports+ which is integrated into Sky TV, streaming service NOW and the Sky Sports app

www.skysports.com

Wimbledon: All England club to replace all 300 line judges after 147 years with electronic system next year​

Wimbledon will replace all 300 line judges from next year with AI technology; Watch all the ATP and WTA action on Sky Sports Tennis and Sky Sports+ which is integrated into Sky TV, streaming service NOW and the Sky Sports app

Wednesday 9 October 2024 18:33, UK

Image:Wimbledon have confirmed they will replace all 300 line judges with an electronic system after 147 years
https://www.facebook.com/sharer/sha...er 147 years with electronic system next year

Wimbledon will replace all 300 line judges with artificial intelligence and an electronic system from next year, removing human officials for the first time in the championship's 147-year history.

The All England Club has confirmed all 18 match courts will be installed with automated electronic line calling (ELC).

The system will be the same as the one used at the US Open since 2020, where automated voice calls are played out within a tenth of a second of the ball landing.

As a result of the change, it is expected that Wimbledon's Hawk-Eye challenge system - brought into use in 2007 - where players could review calls made by the line judges will be removed.

In recent years, 300 line judges aged between 18 and 80 have been used at Wimbledon to cover 650 matches over a two-week period.

Image:All 18 match courts at next year’s tournament will use an electronic line calling system

"The decision to introduce Live Electronic Line Calling at The Championships was made following a significant period of consideration and consultation," said Sally Bolton, chief executive at the All England Lawn Tennis Club.

"Having reviewed the results of the testing undertaken at The Championships this year, we consider the technology to be sufficiently robust and the time is right to take this important step in seeking maximum accuracy in our officiating. For the players, it will offer them the same conditions they have played under at a number of other events on tour.

"We take our responsibility to balance tradition and innovation at Wimbledon very seriously. Line umpires have played a central role in our officiating set-up at The Championships for many decades and we recognise their valuable contribution and thank them for their commitment and service."

The French Open, which has relied solely on human officials, is now the outlier, although it may well follow suit, with all ATP Tour events to feature the technology from next year.

'Line judge change inevitable'​

Image:John McEnroe arguing over a line-call with the umpire will be consigned into the history books

Sky Sports Tennis' commentator Jonathan Overend:

"It is sad, it's quite quaint having the line judges there in their regalia at the side of the court and with the calls going on we'd never have had the John McEnroe 'You cannot be serious' if the robots had guarded the line 40 years ago.

"But it is inevitable because what's happening on the tennis tours next year is automatic line calling is coming in as a matter of course, so the robots will be guarding the lines.

"It's been a success in the tournaments this year, it really does work. People thought it wouldn't work as well as it has done but it undoubtedly works so the grand slams can't be left behind.

"It's already in place at the Australian Open and the US Open so the big question was what were Wimbledon going to do. It will be a huge change and it will mean that the battalion of line judges will have to find something else to do for a fortnight which is a shame because for so many of them that has been their dream moment of the year for decades in some cases."

Wimbledon also announce change to final weekend schedule​

Image:Spain's Carlos Alcaraz became 2024's Wimbledon singles champion

Wimbledon have also announced they are adjusting the schedule to their final weekend.

For 2025, the men's and women's doubles finals will start at 1pm, followed by the women's and men's singles finals at 4pm on the second Saturday and Sunday respectively.

The singles semi-finals schedules will remain the same, while the slot for the mixed doubles final will also remain as the third match on the second Thursday.

Bolton said: "We have adjusted the provisional schedule for the final weekend of The Championships with the ambition of improving the experience for all involved.

"The doubles players competing in the finals will have increased certainty over their schedule and fans will enjoy each day's play as it builds towards the crescendo of the ladies' and gentlemen's singles finals, with our champions being crowned in front of the largest possible worldwide audience."

 

[bnew]

Amazon's new warehouses will employ 10x as many robots | TechCrunch

At its Delivering the Future event Wednesday, Amazon announced plans for new robot-powered delivery warehouses. The first “next-generation fulfillment

techcrunch.com

Amazon’s new warehouses will employ 10x as many robots​

Brian Heater

8:05 AM PDT · October 9, 2024

At its Delivering the Future event Wednesday, Amazon announced plans for new robot-powered delivery warehouses. The first “next-generation fulfillment center” is located in Shreveport, Louisiana. The 3-million-square-foot warehouse spans five floors, constituting the rough equivalent of 55 football fields, per the company.

The site represents the culmination of Amazon’s work in robotics, which dates back more than a decade to its 2012 Kiva acquisition. The retail giant’s approach has largely revolved around incorporating robots into existing worklfows, so as to not disrupt regular operations. The new model looks to bring a more ground-up greenfield approach to robotics and AI.

Amazon has yet to announce specific figures in terms of robots deployed, only that it will bring 10x that of a standard fulfillment center. We do know, however, that the company already has nearly a million robotic systems deployed in centers across the U.S.

Along with the Kiva-style autonomous mobile robots (AMRs) and inventory robotic arms Robin, Cardinal, and Sparrow, the company is deploying Sequoia, which it refers to as “a state-of-the-art multilevel containerized inventory system that makes it faster and safer for employees to store and pick goods. In our next-generation facility, Sequoia can hold more than 30 million items.”

This version is 5x the size of the first Sequoia inventory system the company deployed in Houston around this time last year.

Likely the August hiring of Covariant founders Pieter Abbeel, Peter Chen, and Rocky Duan will play a pivotal role in deploying AI across the system. The Louisiana fulfillment center will prove a key test for these robots to work in tandem. That’s been a difficult nut to crack, as hardware-agnostic platforms and communications between robotic systems was something of a pipe dream until fairly recently.

Amazon is quick to note that there will still be humans in the loop with these systems. It says the Louisiana location will employ 2,500 people once fully ramped up.

 

[bnew]

Spoiler: full text & large images

1/8
@DKThomp
Fascinating paper (and post!) on AI in medicine

- cardiologists working with AI said the machine's diagnosis, triage and management were equal or better than human cardiologists in most areas

- the paper's summary comes with a generative AI podcast explaining the results

[Quoted tweet]
One of the most promising areas for the application of AI in medicine is scaling specialty expertise. There simply aren't enough specialist doctors to care for everyone in need. We believe AI can help.

As a first step towards that goal, we worked with the amazing Google medical AI team to tune and test their conversational agent AMIE in the setting of Stanford's Center for Inherited Cardiovascular Disease.

Unlike many medical studies of LLMs, we completed our testing not with curated cases or exam questions but real-world medical data presented in exactly the way we receive it in clinic.

Data was in the form of reports derived from multi-modal data sources including medical records, ecgs, stress tests, imaging tests, and genomic data. AMIE was augmented by web search and self-critique capabilities and used chain-of-reasoning strategies fine-tuned on data from just 9 typical patients.

What did we find?

1. Overall, AMIE responses on diagnosis, triage and management were rated by specialty cardiologists as equivalent to or better than those of general cardiologists across 10 domains.

2. Access to AMIE's responses improved the general cardiologists' responses in almost two thirds of cases.

3. Qualitative data suggested that the AI and human approaches were highly complementary with AMIE judged thorough and sensitive and general cardiologists judged concise and specific.

In conclusion, our data suggest that LLMs such as AMIE could usefully democratize subspecialty medical expertise augmenting general cardiologists' assessments of inherited cardiovascular disease patients.

Paper: arxiv.org/abs/2410.03741
Generative podcast describing the paper (!): shorturl.at/rdKZn
Stanford Center for Inherited Cardiovascular Disease: med.stanford.edu/familyheart
AMIE: arxiv.org/abs/2401.05654

Congrats to @DrJackOSullivan and @taotu831 for leading the charge on this work as well as the @StanfordDeptMed team and the amazing folks @Google led by @alan_karthi and @vivnat

2/8
@Jill992004231
AI is coming at us very, very quickly.



3/8
@ian_sportsdev
So? Tech is easy, politics is hard



4/8
@fl_saloni
Cardiologists are the gatekeepers for adoption of this AI and most (NOT all) will create mistrust in AI to preserve their status and comp. How do you overcome this



5/8
@Spear_Owl
AI+DR>DR



6/8
@mario_anchor
The question is will specialty caregivers allow this. Their lobbying apparatus has already created an artificial doctor shortage to prop up wages.



7/8
@EscoboomVanilla
Also this!

Wimbledon staff left devastated after decision to break 147-year tradition and put 300 jobs at risk



8/8
@PatrickPatten8
I think medical and education is where AI will have the biggest impact. Hopefully America rethinks what an educated population looks like, cause right now 47% think fascism is a good idea... hopefully we can do better.




To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

 

[GnauzBookOfRhymes]

AI wage discrimination. It's a cold world out here. AI used to adjust your pay based on how desperate it determines you are.

Uber Boss Makes Shocking Admission Over ‘Algorithmic Wage Discrimination’ | Novara Media

Exclusive: Uber’s CEO said it would get better at paying drivers ‘based on behavioural patterns’ - a controversial policy which the company has long denied.

novaramedia.com

Uber’s CEO admitted that the company changes its drivers’ pay based on their “behavioural patterns”, after years of denials of the controversial policy which has been called “algorithmic wage discrimination”.

The Silicon Valley gig economy giant has been accused for years of gathering data from drivers and using it to decide what it pays them. Academics and trade unions claim that the policy is used to reduce pay by stealth while undermining laws on equality at work.

The claim is key to rising dissatisfaction among Uber’s drivers and riders internationally over declining and unpredictable pay rates, with strikes set for the UK, Canada and the US this week.

Uber has always denied the claim.

However, in a conference call with top investors on 7 February, CEO Dara Khosrowshahi was asked specifically about Uber’s “upfront fares” policy, and responded: “I think what we can do better is targeting different trips to different drivers based on their preferences, or based on behavioural patterns that they are showing us.

“That is really the focus going forward: Offering the right trip, at the right price to the right driver.”

 

[bnew]

https://www.pcmag.com/news/tiktok-lays-off-hundreds-of-staff-to-replace-them-focus-on-ai

TikTok Lays Off Hundreds of Staff—to Replace Them With AI​

TikTok is reportedly ditching more of its human moderators to focus on AI-powered moderation instead.

https://www.pcmag.com/authors/kate-irwin

By Kate Irwin

October 11, 2024
https://xenforo.com/community/javascript:void(0)

(Credit: Victollio/Shutterstock.com)

Is AI taking human jobs? It already has been—and is now being used to effectively replace some human moderators at TikTok.

TikTok has confirmed it's laying of "hundreds" of employees globally as part of a broader shift toward AI-powered content moderation. "We're making these changes as part of our ongoing efforts to further strengthen our global operating model for content moderation," a TikTok spokesperson said in a statement to Reuters.

An estimated 500 TikTok employees in Malaysia are being laid off, according to the video platform's parent company ByteDance. Laid-off staff were notified of their terminations by email on Wednesday, according to some of the employees. TikTok is also looking to further consolidate some of its regional operations in the coming months—and previously laid off staff in marketing and operations earlier this year.

TikTok already uses a mix of human and automated content moderation, where AI scans uploaded videos for potential nudity, violence, or other material that could violate TikTok's rules. If a user appeals a ruling that restricts or flags their content, a human moderator may review it.

But TikTok's human moderators have never had it easy. In 2022, The Bureau of Investigative Journalism reported that TikTok's human moderators around the world are only paid the equivalent of about $1.80 an hour or $10 a day—and are regularly required to review videos containing "murder, suicide, pedophili[a], pornographic content, accidents, [or] cannibalism." Many of the workers live in developing countries like Mexico, Malaysia, or Colombia.

TikTok moderators told the outlet that any psychological support is just for show, and that moderators are heavily surveilled. If they don't review a minimum number of videos a month, they may lose up to a quarter of their salary as deducted bonus pay. One moderator explained that they have to review 900 videos a day, and only watching 700 videos a day is considered "work avoidance."

At the time, a TikTok rep told the Bureau: "We strive to promote a caring working environment for our employees and contractors. Our Trust and Safety team partners with third party firms on the critical work of helping to protect the TikTok platform and community, and we continue to expand on a range of wellness services so that moderators feel supported mentally and emotionally."

The report suggested TikTok may be using a mix of human and AI reviews because, at that time, human moderators were seen as cheaper and more accurate than AI.

TikTok did not immediately respond to a request for comment.

 

[bnew]

 

[bnew]

Spoiler: thread continued

Spoiler: full text & large images

1/9
@amazonnews
Reporting live from Delivering the Future, where we're sharing our latest and greatest innovations in delivery technology, robotics, and /search?q=#AI.



2/9
@amazonnews
Wondering what our fulfillment center of the future looks like? Located in Shreveport, Louisiana, this state-of-the-art facility is powered by the latest in /search?q=#robotics and /search?q=#AI. Take a look. Amazon unveils the next generation of fulfillment centers powered by AI and 10 times more robotics

3/9
@amazonnews
Green means grab, and go – and that’s exactly how Amazon’s Vision Assisted Package Retrieval (VAPR), a new /search?q=#AI-powered solution, is making drivers’ lives easier. How?

VAPR uses a combo of advanced LED light projectors, overhead cameras, and ambient lighting to locate and scan barcodes inside a delivery van, and cross-references them with the route itinerary.

VAPR will automatically project a green “O” on all packages that will be delivered at that stop and a red “X” on all other packages. New AI tech in Amazon vans spotlights packages to save drivers effort and time

4/9
@amazonnews
We're happy to share that we're continuing to make huge strides in making our packaging easier to recycle.

This year, we removed plastic air pillows from deliveries from our global fulfillment centers and have retrofitted more than 120 machines that previously made plastic bags to now make paper bags in the US - helping us to avoid 130 million plastic bags this year. How Amazon is improving packaging and boosting sustainability

5/9
@amazonnews
We're making your grocery shopping simpler, faster, and more affordable than ever before.

Here's the latest.

1⃣ In Phoenix, Same-Day and Overnight Delivery of groceries and general merchandise in one order: No more sweating last-minute shopping needs!

2⃣ You’ll see broader selection from other stores when shopping with Amazon Fresh Online in select cities: Find organic products from @WholeFoods, your favorite brand names, and low-cost snacks from Amazon Fresh, plus general merchandise like socks from Amazon.com, delivered in a single order to your door.

3⃣ Pick up national brands and household goods while shopping at Whole Foods Market: At a location in Pennsylvania, we’re building Amazon’s first-ever automated micro fulfillment center stocked with popular Amazon.com and Amazon Fresh items that customers can pick up, have delivered, or even order in-store and grab when they check out.

Learn more. 3 new ways Amazon is offering more choices while making grocery shopping faster and easier

6/9
@amazonnews
Great news for all Amazon Pharmacy customers: We plan to open Same-Day pharmacies in 20 new cities across the U.S. in 2025.

That means nearly half the U.S. will be able to get medications delivered to their door within hours. Amazon Pharmacy plans to expand Same-Day Delivery of medications to nearly half the US in 2025



https://video.twimg.com/ext_tw_video/1844035014566051840/pu/vid/avc1/1280x720/sPKxKm-_8tFBdK-J.mp4

7/9
@amazonnews
.@climatepledge, our corporate venture program for a net-zero future, is investing in 3 new startups bringing innovative tech and fresh thinking to the climate crisis.

Molg enables circular manufacturing with robotics and design to help tackle the growing e-waste problem.

@Paebbl is turning CO2 into carbon-storing building materials, transforming the built environment into a permanent carbon sink.

14Trees is accelerating the production and delivery of lower-carbon buildings with its 3D printing technology.

Amazon’s Climate Pledge Fund adds investments in startups focused on e-waste, carbon storage, and lower-carbon buildings



8/9
@amazonnews
Our robots are helping us get to the next level, powering the next gen fulfillment facility at Amazon. How?

By using /search?q=#AI to ensure the right products are in fulfillment centers near customers, maximizing efficiency for our employees, lifting heavy objects for us, and so much more.

Get to know Proteus, Hercules, and Sequoia, just to name a few! Meet the 8 robots powering your Amazon package deliveries

9/9
@digitalbeacon_
/search?q=#amazon /search?q=#robots # fulfillment
@memdotai mem it
@readwise save thread




To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

 

[bnew]

Spoiler: thread continued

Spoiler: full text & large images

1/9
@deedydas
Got Claude Computer Use to "build a Timer app" from scratch!

—downloaded React + deps
—wrote the code
—opened local server (in a Docker VM)
—fixed its own styling
—TESTED it fully
—restyled it like an iPhone
—fixed a syntax error

Software engineering will not be the same.

1/4



https://video.twimg.com/amplify_video/1849118641935253504/vid/avc1/1280x720/XX-p0M2VzMHfLHR1.mp4

2/9
@deedydas
Observations:
— The new Claude Sonnet has been RLHF'd to not do a bunch of things like fill forms, solve captchas or login to pages.
— It's still slow. This video took ~7mins in realtime, was sped up 16x
— New abilities I hadn't seen before: point + click, bulk copy strings

2/4



3/9
@deedydas
— This demo loads Ubuntu in a Docker image. UI is nowhere like the demo videos, and its a bit clunky. 120s timeout, you can't tell the progress of longer running bash tasks.
— Managed to build backend + frontend apps
— Can learn how to use APIs from the internet

3/4



4/9
@deedydas
So much interesting stuff can be built on top of this for the first time ever.

If you like hacking on this, apply quickly to join us on Nov 2 for the Menlo x Anthropic Builder Day 2024 (sadly Computer Use can't fill this form for you yet)!

Menlo/Anthropic Builder Day 2024

4/4



5/9
@dmsimon
Wow, a timer.



6/9
@pathuglife
Great demo Deedy!

One observation I’ve made while working with browsers is that screenshots are often taken and processed even when pages are still loading, resulting in unnecessary token usage and computation costs. By ensuring that screenshots are only captured after the page has fully loaded, we could significantly reduce compute overhead and improve efficiency.



7/9
@typesteady


[Quoted tweet]
"I built this entirely with AI"

Yes, I can tell.


8/9
@HadijPk
good demo. I feel like Claude is on to something very interesting. I can't wait it to replace old RPA in legacy enterprises



9/9
@FeatureCrewPod
Its better at everyday tasks than coding imp

[Quoted tweet]
New #AI Agent from @AnthropicAI can now...
Delete emails
Manage files
Try to draw
Watch the full video:


https://video.twimg.com/amplify_video/1848941581199400960/vid/avc1/1920x1080/NGOul_PPaOzxRNk9.mp4


To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

 

[bnew]

Google CEO says more than a quarter of the company's new code is created by AI

Google is doubling down on AI internally to make its business more efficient.

www.businessinsider.com

Google CEO says more than a quarter of the company's new code is created by AI​

Hugh Langley

Google

• More than a quarter of new code at Google is made by AI and then checked by employees.
  
• Google is doubling down on AI internally to make its business more efficient.
  
• Business Insider previously reported that Google launched an internal AI model named "Goose."

Google is all in on AI — both inside and outside the company.

More than a quarter of new code created at Google is generated by AI, CEO Sundar Pichai said on Tuesday during the company's Q3 earnings call.

Pichai said using AI for coding was "boosting productivity and efficiency" within Google. After the code is generated, it's then checked and reviewed by employees, he added.

"This helps our engineers do more and move faster," Pichai said. "I'm energized by our progress and the opportunities ahead, and we continue to be laser-focused on building great products."

Business Insider reported in February that the company had launched a new internal AI model named "Goose" to help employees code and build products.

Goose was trained on "25 years of engineering expertise at Google," according to internal documents seen by BI.

The new data from Pichai will surely have some employees wondering whether they're coding themselves out of a job, while other employees say AI has already transformed their work. Company leaders have previously promised that AI isn't taking Googlers' jobs (yet), but the over-25% figure is striking and underscores the benefits of improving this technology.