From Digital Age to Nano Age. WorldWide.

Tag: preview

Robotic Automations

Pinecone launches its serverless vector database out of preview | TechCrunch


Pinecone, the vector database startup founded by Edo Liberty, the former head of Amazon’s AI Labs, has long been at the forefront of helping businesses augment large language models (LLMs) with their own data. Most recently, though, the company completely rearchitected its product to launch Pinecone Serverless, which frees its customers from having to think […]

© 2024 TechCrunch. All rights reserved. For personal use only.


Software Development in Sri Lanka

Robotic Automations

Google's Gemini Pro 1.5 enters public preview on Vertex AI | TechCrunch


Gemini 1.5 Pro, Google’s most capable generative AI model, is now available in public preview on Vertex AI, Google’s enterprise-focused AI development platform. The company announced the news during its annual Cloud Next conference, which is taking place in Las Vegas this week.

Gemini 1.5 Pro launched in February, joining Google’s Gemini family of generative AI models. Undoubtedly its headlining feature is the amount of context that it can process: between 128,000 tokens to up to 1 million tokens, where “tokens” refers to subdivided bits of raw data (like the syllables “fan,” “tas” and “tic” in the word “fantastic”).

One million tokens is equivalent to around 700,000 words or around 30,000 lines of code. It’s about four times the amount of data that Anthropic’s flagship model, Claude 3, can take as input and about eight times as high as OpenAI’s GPT-4 Turbo max context.

A model’s context, or context window, refers to the initial set of data (e.g. text) the model considers before generating output (e.g. additional text). A simple question — “Who won the 2020 U.S. presidential election?” — can serve as context, as can a movie script, email, essay or e-book.

Models with small context windows tend to “forget” the content of even very recent conversations, leading them to veer off topic. This isn’t necessarily so with models with large contexts. And, as an added upside, large-context models can better grasp the narrative flow of data they take in, generate contextually richer responses and reduce the need for fine-tuning and factual grounding — hypothetically, at least.

So what specifically can one do with a 1 million-token context window? Lots of things, Google promises, like analyzing a code library, “reasoning across” lengthy documents and holding long conversations with a chatbot.

Because Gemini 1.5 Pro is multilingual — and multimodal in the sense that it’s able to understand images and videos and, as of Tuesday, audio streams in addition to text — the model can also analyze and compare content in media like TV shows, movies, radio broadcasts, conference call recordings and more across different languages. One million tokens translates to about an hour of video or around 11 hours of audio.

Thanks to its audio-processing capabilities, Gemini 1.5 Pro can generate transcriptions for video clips, as well, although the jury’s out on the quality of those transcriptions.

In a prerecorded demo earlier this year, Google showed Gemini 1.5 Pro searching the transcript of the Apollo 11 moon landing telecast (which comes to about 400 pages) for quotes containing jokes, and then finding a scene in movie footage that looked similar to a pencil sketch.

Google says that early users of Gemini 1.5 Pro — including United Wholesale Mortgage, TBS and Replit — are leveraging the large context window for tasks spanning mortgage underwriting; automating metadata tagging on media archives; and generating, explaining and transforming code.

Gemini 1.5 Pro doesn’t process a million tokens at the snap of a finger. In the aforementioned demos, each search took between 20 seconds and a minute to complete — far longer than the average ChatGPT query.

Google previously said that latency is an area of focus, though, and that it’s working to “optimize” Gemini 1.5 Pro as time goes on.

Of note, Gemini 1.5 Pro is slowly making its way to other parts of Google’s corporate product ecosystem, with the company announcing Tuesday that the model (in private preview) will power new features in Code Assist, Google’s generative AI coding assistance tool. Developers can now perform “large-scale” changes across codebases, Google says, for example updating cross-file dependencies and reviewing large chunks of code.




Software Development in Sri Lanka

Robotic Automations

A CES 2024 preview, 23andMe victim blaming and MIT's obesity-fighting pill | TechCrunch


Welcome, folks, to Week in Review (WiR), TechCrunch’s regular newsletter that recaps the week in tech that was. Hope the holidays were restful for those who observed them. We at TC, for our parts, are gearing up for an eventful next week at CES in Las Vegas — while keeping an eye on the news cycle, as ever.

In this edition of WiR, we spotlight Brian’s CES 2024 preview, 23andMe blaming victims for its data breach, GitHub making Copilot Chat generally available and Frontdesk laying off its entire staff. Also on the agenda are spiders and body butter, Fidelity marking down X’s valuation, Meta cutting the price of the Quest 2 and MIT scientists’ vibrating obesity pill.

It’s a lot to get through, so we won’t delay. But first, a reminder to sign up here to receive WiR in your inbox every Saturday if you haven’t already done so.

Most read

CES 2024: Brian has a thorough roundup of what to expect at CES 2024, including — but not limited to — generative AI, robotics, TVs, cars, smartphones, and health tech. He writes that he’s optimistic about the show overall, particularly in light of the consumer electronics industry’s move to more decentralized manufacturing and the quality of startup pitches that’ve come in so far.

Your fault, not ours: Facing over 30 lawsuits from victims of a data breach implicating ~6.9 million customers, 23andMe is now deflecting blame to attempt to absolve itself of any responsibility. In a letter, the genetic testing company says that users “negligently recycled and failed to update their passwords following these past security incidents, which are unrelated to 23andMe.”

Copilot Chat launches: GitHub has rolled out Copilot Chat, a ChatGPT-like programming-centric chatbot, in general availability for all paying Copilot users and free for verified teachers, students and maintainers of certain open source projects. The chatbot’s powered by GPT-4, OpenAI’s flagship generative AI model, and fine-tuned specifically for dev scenarios.

Frontdesk implodes: Mary Ann writes that Frontdesk, a startup that managed more than 1,000 furnished apartments across the U.S., laid off its entire 200-person workforce Tuesday after attempts to raise more capital failed. Frontdesk CEO Jesse DePinto said that Frontdesk would be filing for a state receivership, an alternative to bankruptcy, according to TechCrunch’s sources.

Spiders and body butter: Sol de Janeiro’s Delícia Drench Body Butter went viral on social media after users claimed they were hunted, bitten and (unsuccessfully) courted by wolf spiders when they applied the moisturizer, thanks to the alleged inclusion of chemicals that spiders find sexually arousing. But Sol de Janeiro — and independent experts — tell TechCrunch that there’s no merit to the rumors.

X’s valuation falls . . . again: Mutual fund company Fidelity has marked down its investment in X Holdings, the parent company of X (formerly Twitter), by 71.5% from the original valuation of shares, reports Ivan. Fidelity spent $19.2 million to acquire a stake in X back in October 2022.

Quest 2 discounted: Months after Meta launched the Quest 3, the company is slashing prices for the VR headset’s predecessor, the Quest 2, by $50. The 128GB version drops from $299 to $249 and the 256GB version drops from $349 to $299 — with plenty of accessories on sale to boot.

Vibrating the fat away: Brian writes about an MIT team’s new obesity-fighting, pill-like vibrating capsule, which is designed to send signals to the brain to simulate the sensation of being full. Early tests are promising — giving animals the pill 20 minutes before eating reduced their consumption by around 40% — but the capsule is a long way from human trials.

Audio

In need of new podcasts to fill out your rotation? Don’t panic — TechCrunch has you covered.

On a throwback episode of Equity, Morgan interviewed Shruti Dwivedi — the co-founder and CEO of health tech startup Duly, which is focused on simplifying and personalizing contraception for young women in India and beyond — at TechCrunch Disrupt 2023. The pair talked about the stigma around contraception, cultural roadblocks Duly faces and what’s next for the startup.

Meanwhile, Found went Down Under with Rebecca, who spoke with Alex Zaccaria, the co-founder and CEO of Australia-based Linktree. The two chatted about how the startup scaled the freemium model to grow the now-massive social media reference landing page business.

And on Chain Reaction, Jacquelyn dove back into the latest developments on spot bitcoin ETF applications in the U.S. as anticipation builds. Fred Thiel, the CEO of Marathon Digital Holdings, a digital asset technology company and the largest publicly traded bitcoin mining firm, joined the episode to talk crypto shop.

TechCrunch+

TC+ subscribers get access to in-depth commentary, analysis and surveys — which you know if you’re already a subscriber. If you’re not, consider signing up. Here are a few highlights from this week:

Another alleged cool superconductor: Tim has the story on the latest team of scientists who claim to have discovered a near-room-temperature superconductor. He’s not convinced that the paper, which hasn’t been peer-reviewed, will stand up to scientific scrutiny; time will tell.

Crypto losses decline: While malicious actors continue to hack the crypto industry for a cash grab, the dollar amount is down substantially — 51% — compared to the previous year, Jacquelyn writes.

The coming copyright challenges: When news broke last year that AI heavyweight OpenAI and Axel Springer had reached a financial agreement and partnership, it seemed to bode well for harmony between folks who write words and tech companies that use them to help create and train AI models. But perhaps it doesn’t, Alex posits.


Software Development in Sri Lanka

Back
WhatsApp
Messenger
Viber