Connect with us

Technology

Letta, one of UC Berkeley’s most anticipated AI startups, has just come out of stealth

Published

on

Letta, one of UC Berkeley’s most anticipated AI startups, has just come out of stealth

A startup called Letta has just emerged from stealth with tech that helps AI models remember users and conversations. Created in UC Berkeley’s famed Labs startup factory, it also announced $10 million in seed money led by Felicis’ Astasia Myers, at a $70 million post-money valuation. 

Letta is also backed by a who’s who of angel investors in AI, like Google’s Jeff Dean, Hugging Face’s Clem Delangue, Runway’s Cristóbal Valenzuela, and Anyscale’s Robert Nishihara, among others.

Founded by Berkeley PhD students Sarah Wooders and Charles Packer, this is a highly anticipated AI startup launch. That’s because it’s a child of Berkeley’s Sky Computing Lab and is the commercial entity of the popular MemGPT open source project.

Berkeley’s Sky Computing Lab, led by acclaimed professor and Databricks co-founder Ion Stoica, is the descendent of RISELab and AMPLab, which spawned such companies as Anyscale, Databricks, and SiFive. Sky Lab, in particular, birthed numerous popular open source large language model (LLM) projects like the Gorilla LLM, vLLM, and the LLM structured language SGLang.

Advertisement

“A ton of projects very quickly, within a year’s time frame, came out of the lab. Just people sitting next to us,” described Wooders. “So it was kind of an incredible time.”

MemGPT is one such project and is such a hot commodity that it actually went viral before it even launched. 

“Someone scooped us,” Packer told TechCrunch. The founders had posted a whitepaper on Thursday, October 12, 2023, and planned to release a more in-depth paper and the code to GitHub the following Monday. Some random person found the paper, posted it to Hacker News on Sunday, and it “went viral on Hacker News before we had a chance to properly release the code, release the paper, or, like, do a tweet thread or anything like that,” he said. 

The reason for the excitement was that MemGPT mitigates a pernicious problem for LLMs: In their native form, models like ChatGPT are stateless, meaning they don’t store historical data in long-term memory. This is problematic for AI apps that depend on getting to know and learn from a user over time — everything from customer support bots to healthcare symptom-tracking apps. MemGPT manages data and memory so that AI agents and chatbots can remember previous users and conversations.

Advertisement

The post on the paper stayed atop Hacker News, the popular site for programmers run by Y Combinator, for 48 hours, Packer recounted. So he spent his weekend and the next few days answering questions on the site while trying to get the code ready to be released. Once the project was available on GitHub, a link to it went viral on Hacker News, again. YouTube interviews and tutorials, Medium posts, 11,000 stars and 1.2K forks on GitHub happened quickly.

VC Felicis’ Myers discovered Wooders and Packer by reading about MemGPT, too, and immediately recognized the tech’s commercial possibilities.

“I saw the paper when it was released,” she told TechCrunch, and she promptly reached out to the founders. “We had an investment theme around AI agent infrastructure and appreciated that a really important component of that was the data and memory management to make these conversational chat bots and AI agents effective.”

The founders still virtually traipsed around Sand Hill Road doing Zoom calls with VCs before going with the one that loved them first.

Advertisement

Meanwhile, Stoica brokered introductions to Dean, Nishihara and other big-name Silicon Valley angels. “A lot of the professors at Berkeley, just as a consequence of being at Berkeley, are very well connected,” Packer recalled about how easy the angel investor process was. “They have their eye on projects out of this lab that are going to be commercialized.”

Competition and the threat of OpenAI o1

While MemGPT is already out in the wild and being used, Letta’s commercial variant, Letta Cloud, is not yet open for business. As of Monday, Letta is accepting requests for beta users. It will offer a hosted agent service that allows developers to deploy and run stateful agents in the cloud, accessible via REST APIs, a programming interface that can maintain state. Letta Cloud will store the long-term data necessary to do so. Letta will also offer developer tools for building AI agents.

With MemGPT, Wooders sees a large span of uses. “I think the number one use case that we see is basically, highly personalized, very engaging chatbots,” she says. But there are also cutting-edge uses like “a chatbot for cancer patients” where patients upload their history and then share ongoing symptoms so the bot can learn and offer guidance over time. 

Worth noting that MemGPT isn’t alone in working on this. LangChain is probably its best known competitor, and it already offers commercial options. The biggest model makers also offer AI agent-making tools as well, like OpenAI’s Assistants API. 

Advertisement

And OpenAI’s new o1 model may make the need to fix state a moot point for its users. As it is a multistep model, it fundamentally must maintain state to some degree in order to “think” and fact check before it replies.

But Wooders, Packer, and Myers see a few key differences to what Letta is offering versus what 800-pound market gorilla OpenAI is doing. Letta claims it will work with any AI model and expects its users to use many of them: OpenAI, Anthropic, Minstrel, their own homegrown models. OpenAI’s tech currently only works with itself. 

More importantly, Letta is using open source MemGPT and jumping firmly into the open source side of the FOSS vs. black box LLM debate, saying open source is a better choice for AI application programmers.

“We are positioning ourselves as the open alternative to OpenAI,” Packer says. “I think it’s actually very, very hard to build very good AI applications, especially when you care about, like hallucination, if you can’t see what’s going on under the hood.”

Advertisement

Source link

Continue Reading
Advertisement
Click to comment

You must be logged in to post a comment Login

Leave a Reply

Technology

Gemini app on Android gets a more simplified homescreen

Published

on

Gemini app on Android gets a more simplified homescreen

Earlier this month, Google redesigned the Gemini homepage on desktop. The redesign removed the “Hello” + greeting and suggestions and now offers a simple “Hello can I help you?” message. The search engine giant has now provided the same on mobile devices. The Gemini app on Android devices has received a new simplified homescreen, which is now more minimalistic than earlier.

Gemini app’s new simplified homescreen replaces the old design with a minimalistic interface

According to a report by 9to5Google, the Gemini app on Android has received a new update, which brings a more simplified homescreen. The new update replaces the old design with a more minimalistic interface. The new UI is designed to prioritize user experience.

Earlier, there was the “Gemini Advanced” logo on top, apart from a handful of suggestions and the “Chats & Gems” list. However, the new homescreen only has the “Hello, [user name]” greeting. It now also includes a new message bubble icon in the top-left corner, which gives you access to the Recent conversations and Gems. Notably, the chat box at the bottom of the screen is unchanged.

As you can see, the new Gemini app interface draws inspiration from the search’s homepage. If you want to access the new Gemini interface, then simply update the app on your Android device. The new design seems to be rolling out gradually, it may not be available in all regions right away.

Advertisement

Google recently also added Gems to Gemini’s mobile app

It’s worth mentioning that Google added Gems to Gemini’s mobile app earlier this month. Gems were originally launched for gemini.google.com at the end of August on the web. However, now they are also available in the Gemini app on Android and iOS devices. Earlier, the “Recent” section was renamed to “Chat & Gems”. However, with the latest homescreen layout update, they are moved to the bubble icon in the top-left corner.

For those unaware, Gems in Gemini allows you to program the AI assistant for a single and repetitive task. Notably, they can “remember a detailed set of instructions to help you save time on tedious, repetitive or difficult tasks”. Google also offers a handful of pre-generated Gems like a learning coach, brainstormer, career guide, and writing editor.

Source link

Advertisement
Continue Reading

Technology

Prime Day deals include three free months of Amazon Music Unlimited for new users

Published

on

Prime Day deals include three free months of Amazon Music Unlimited for new users

Amazon is hoping to get more people hooked on Music Unlimited with a pre-Prime Day offer. Those who haven’t tried the service before can get three months of access for free. The offer is even better for Prime members who haven’t checked out the music-streaming platform. They can use it for four months at no extra cost. Music Unlimited typically costs $11 per month without Prime and $10 per month for Prime members.

The company isn’t leaving existing subscribers entirely out in the cold. Those with an individual Amazon Music Unlimited plan can upgrade to the Family Plan at no extra cost for two months.

Amazon

Advertisement

Amazon Music Unlimited is free for three months to those who haven’t used it before. Prime members who are newcomers to the service get an additional month of access at no extra cost.

$0 at Amazon

Amazon Music Unlimited includes more than 100 million songs, many of which are available in high definition and/or spatial audio. The service also includes ad-free versions of many popular podcasts.

While it may not be the first music streaming service that springs to mind for most folks, Amazon Music Unlimited may be worth checking out. The HD music selection is particularly notable, since Spotify has yet to offer that feature.

Advertisement

Follow @EngadgetDeals on Twitter for the latest tech deals and buying advice in the lead up to October Prime Day 2024.

Source link

Advertisement
Continue Reading

Technology

Together AI promises faster inference and lower costs with enterprise AI platform for private cloud

Published

on

Together AI promises faster inference and lower costs with enterprise AI platform for private cloud

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More


Running AI in the public cloud can presents enterprises with numerous concerns about data privacy and security.

That’s why some enterprises will choose to deploy AI on a private cloud or on-premises environment. Together AI is among the vendors looking to solve the challenges of effectively enabling enterprises to deploy AI in private clouds in a cost effective approach. The company today announced its Together Enterprise Platform, enabling AI deployment in virtual private cloud (VPC) and on-premises environments.

Together AI made its debut in 2023, aiming to simplify enterprise use of open-source LLMs. The company already has a full-stack platform to enable enterprises to easily use open source LLMs on its own cloud service. The new platform extends AI deployment to customer-controlled cloud and on-premises environments. The Together Enterprise Platform aims to address key concerns of businesses adopting AI technologies, including performance, cost-efficiency and data privacy.

Advertisement

“As you’re scaling up AI workloads, efficiency and cost matters to companies, they also really care about data privacy,” Vipul Prakash, CEO of Together AI told VentureBeat. “Inside of enterprises there are also well-established privacy and compliance policies, which are already implemented in their own cloud setups and companies also care about model ownership.”

How to keep private cloud enterprise AI cost down with Together AI

The key promise of the Together Enterprise Platform is that organizations can manage and run AI models in their own private cloud deployment.

This adaptability is crucial for enterprises that have already invested heavily in their IT infrastructure. The platform offers flexibility by working in private clouds and enabling users to scale to Together’s cloud.

A key benefit of the Together Enterprise platform is its ability to dramatically improve the performance of AI inference workloads. 

Advertisement

“We are often able to improve the performance of inference by two to three times and reduce the amount of hardware they’re using to do inference by 50%,” Prakash said. “This creates significant savings and more capacity for enterprises to build more products, build more models, and launch more features.” 

The performance gains are achieved through a combination of optimized software and hardware utilization.

 “There’s a lot of algorithmic craft in how we schedule and organize the computation on GPUs to get the maximum utilization and lowest latency,” Prakash explained. “We do a lot of work on speculative decoding, which uses a small model to predict what the larger model would generate, reducing the workload on the more computationally intensive model.”

Flexible model orchestration and the Mixture of Agents approach

Another key feature of the Together Enterprise platform is its ability to orchestrate the use of multiple AI models within a single application or workflow. 

Advertisement

“What we’re seeing in enterprises is that they’re typically using a combination of different models – open-source models, custom models, and models from different sources,” Prakash said. “The Together platform allows this orchestration of all this work, scaling the models up and down depending on the demand for a particular feature at a particular time.”

There are many different ways that an organization can orchestrate models to work together. Some organizations and vendors will use technologies like LangChain to combine models together. Another approach is to use a model router, like the one built by Martian, to route queries to the best model. SambaNova uses a Composition of Experts model, combining multiple models for optimal outcomes.

Together AI is using a different approach that it calls – Mixture of Agents. Prakash said this approach combines multi-model agentic AI with a trainable system for ongoing improvement. The way it works is by using “weaker” models as “proposers” – they each provide a response to the prompt. Then an “aggregator” model is used to combine these responses in a way that produces a better overall answer.

“We are a computational and inference platform and agentic AI workflows are very interesting to us,” he said. “You’ll be seeing more stuff from Together AI on what we’re doing around it in the months to come.”

Advertisement

Source link
Advertisement
Continue Reading

Technology

Erin and Sara Foster join the stage at Disrupt 2024

Published

on

Consumer, culture, and creators with Erin and Sara Foster at TechCrunch Disrupt 2024

Two of the industry’s most famous sisters, Erin and Sara Foster, sit down alongside business partner Phil Schwarz at TechCrunch Disrupt 2024 to talk about consumer investing, culture curation, and what it means to be a creator in the modern age. 

Moderated on the Disrupt main stage, the conversation will surround the creator economy, consumer investing, and so much more.

Erin and Sara Foster are the definitions of a multi-hyphen duo. 

They are the founders of Oversubscribed Ventures, which invests in consumer brands and tech (investments include the diaper brand Kudos). Sara is an angel investor, too, with a portfolio that includes the swimwear brand Summersalt. The sisters are also the founders of the clothing and lifestyle brand Favorite Daughter, which is sold at luxury retailers Moda Operandi and Saks. There’s also an upcoming Netflix show called “Nobody Wants This,” created by Erin and starring Kristen Bell and Adam Brody, playing a podcaster and a rabbi who fall in love. 

Advertisement

Sara and Erin themselves have a podcast called The World’s First Podcast, which breaks down topics of love and self-improvement. 

In other words, there will be much to talk about. 

And of course, not to forget Phil Schwarz. Schwarz is a partner at the early-stage firm Corazon Capital. He helped Sara and Erin Foster create Oversubscribed Ventures and serves as an adviser on its investment committee. He was the CMO of Tinder and Head of Growth at Match Group and co-founded the Kaplan/Techstars EdTech Accelerator.

In other words, even more to talk about!

Advertisement

Don’t miss out on this game-changing conversation! Be part of Disrupt 2024 with 10,000 fellow attendees, including startup, tech, and VC leaders, at Moscone West, San Francisco, October 28–30. Grab your tickets now to unlock savings of up to $600 during Ticket Reboot Week.

Source link

Continue Reading

Technology

The UK helped usher in the coal era — now its closing its last remaining plant

Published

on

The UK helped usher in the coal era — now its closing its last remaining plant

By the end of the month, the United Kingdom is set to shutter its last operating coal-fired power plant.

The Ratcliffe-on-Soar power station in Nottinghamshire is slated to close on September 30th, marking the end of coal power in the UK. It’s turning the page on an era of dirty energy that the UK helped usher in globally and now has to leave behind to meet climate goals.

The coal power plant was first commissioned in 1967, and received its last coal delivery in June. The 2,000 megawatt-capacity plant can produce enough electricity for some two million homes, according to the energy company Uniper that owns it. Over its decades in operation, it’s generated enough energy to make more than 21 trillion cups of tea, or roughly 1 billion cups per day, Uniper says.

Uniper announced plans to eventually produce hydrogen at the site instead of coal power

Advertisement

Still, it’ll reportedly take two years to decommission the plant after it shuts down, requiring 125 staff to stay on board until then. Uniper announced plans last year to eventually produce hydrogen at the site instead of coal power. Unlike fossil fuels, hydrogen fuel doesn’t create planet heating carbon dioxide pollution when burned.

The caveat is that most hydrogen today is made through a process called steam methane reforming, which still leads to greenhouse gas emissions. A cleaner way to make hydrogen is through electrolysis, splitting water molecules using renewable energy. Uniper says it’s interested in producing hydrogen through electrolysis at the former coal plant, reaching a capacity of 500 megawatts by the end of the decade. After redeveloping the site to produce low carbon energy, Uniper says it could create up to 8,000 jobs.

Coal still makes up more than a third of the world’s electricity mix, generating more power and greenhouse gas emissions than any other fuel. But coal-fired power generation has dropped dramatically in the UK since 2012, falling from nearly 40 percent to just 1 percent of the UK’s electricity mix by 2023. Greenhouse gas emissions from the power sector have fallen 74 percent over the same period of time as wind and solar replace coal.

The UK passed a Climate Change Act in 2008 that set it on a path to ditch coal on the way to reaching a target of net zero carbon dioxide emissions by 2050. Once the Ratcliffe-on-Soar plant closes, the UK will become the first advanced economy and Group of 7 nation to phase out coal power, according to climate change think tank E3G.

Advertisement

Source link

Continue Reading

Technology

I want one! MSI quietly debuts sub-1kg laptop with a killer price tag and exceptional performance — prestige 13 AI+ will appeal to businesses and no, don’t get put off by its lack of RAM upgradability

Published

on

I want one! MSI quietly debuts sub-1kg laptop with a killer price tag and exceptional performance — prestige 13 AI+ will appeal to businesses and no, don't get put off by its lack of RAM upgradability

MSI has quietly launched the Prestige 13 AI+ Evo A2VM, a super-lightweight laptop aimed at business professionals and creatives alike.

First unveiled at IFA 2024, the device is powered by up to an Intel Core Ultra 9 288V processor (offering up to 48 NPU TOPS and a total of 120 AI TOPS), paired with Intel Arc 140V graphics.

Source link

Advertisement

Continue Reading

Trending

Copyright © 2017 Zox News Theme. Theme by MVP Themes, powered by WordPress.