For OpenAI, o1 represents a step toward its broader goal of human-like artificial intelligence. More practically, it does a better job at writing code and solving multistep problems than previous models. But it’s also more expensive and slower to use than GPT-4o. OpenAI is calling this release of o1 a “preview” to emphasize how nascent it is.

The training behind o1 is fundamentally different from its predecessors, OpenAI’s research lead, Jerry Tworek, tells me, though the company is being vague about the exact details. He says o1 “has been trained using a completely new optimization algorithm and a new training dataset specifically tailored for it.”

OpenAI taught previous GPT models to mimic patterns from its training data. With o1, it trained the model to solve problems on its own using a technique known as reinforcement learning, which teaches the system through rewards and penalties. It then uses a “chain of thought” to process queries, similarly to how humans process problems by going through them step-by-step.

At the same time, o1 is not as capable as GPT-4o in a lot of areas. It doesn’t do as well on factual knowledge about the world. It also doesn’t have the ability to browse the web or process files and images. Still, the company believes it represents a brand-new class of capabilities. It was named o1 to indicate “resetting the counter back to 1.”

I think this is the most important part (emphasis mine):

As a result of this new training methodology, OpenAI says the model should be more accurate. “We have noticed that this model hallucinates less,” Tworek says. But the problem still persists. “We can’t say we solved hallucinations.”

Google is excelling again - as the whole "uncensored" Big-Tech IT now.

The short summary is that for nearly a year, Google was hiding Proton Mail from search results for queries such as ‘secure email’ and ‘encrypted email’. This was highly suspicious because Proton Mail has long been the world’s largest encrypted email provider.

cross-posted from: https://lemmy.zip/post/22604748

The Vision Pro uses 3D avatars on calls and for streaming. These researchers used eye tracking to work out the passwords and PINs people typed with their avatars.

Archived version: https://web.archive.org/web/20240912100207/https://www.wired.com/story/apple-vision-pro-persona-eye-tracking-spy-typing/

You might sideload an Android app, or manually install its APK package, if you're using a custom version of Android that doesn't include Google's Play Store. Alternately, the app might be experimental, under development, or perhaps no longer maintained and offered by its developer. Until now, the existence of sideload-ready APKs on the web was something that seemed to be tolerated, if warned against, by Google.

This quiet standstill is being shaken up by a new feature in Google's Play Integrity API. As reported by Android Authority, developer tools to push "remediation" dialogs during sideloading debuted at Google's I/O conference in May, have begun showing up on users' phones. Sideloaders of apps from the British shop Tesco, fandom app BeyBlade X, and ChatGPT have reported "Get this app from Play" prompts, which cannot be worked around. An Android gaming handheld user encountered a similarly worded prompt from Diablo Immortal on their device three months ago.

Google's Play Integrity API is how apps have previously blocked access when loaded onto phones that are in some way modified from a stock OS with all Google Play integrations intact. Recently, a popular two-factor authentication app blocked access on rooted phones, including the security-minded GrapheneOS. Apps can call the Play Integrity API and get back an "integrity verdict," relaying if the phone has a "trustworthy" software environment, has Google Play Protect enabled, and passes other software checks.

Graphene has questioned the veracity of Google's Integrity API and SafetyNet Attestation systems, recommending instead standard Android hardware attestation. Rahman notes that apps do not have to take an all-or-nothing approach to integrity checking. Rather than block installation entirely, apps could call on the API only during sensitive actions, issuing a warning there. But not having a Play Store connection can also deprive developers of metrics, allow for installation on incompatible devices (and resulting bad reviews), and, of course, open the door to paid app piracy.

The article mentions more research is needed to confirm if the effect is long-lasting, but personally I'm happy someone may have found a good, practical usecase for LLMs.

There are reports in Registar's comment section that Malaysia didn't only redirect DNS traffic, but took active measures to block VPN, and MITM DoH where Cloudflare's DoH returns local ISP certificate.

In fact, some ISPs like Maxis and Yes were already blocking VPN (I see a lot of complains on Lowyat.net about Maxis blocking VPN, and I was using Yes WiMax and experienced the blocking firsthand. I couldn't connect to PPTP endpoints and L2TP endpoints caused the modem to disconnect from the network and reboot).

They were outright trying a MITM redirect attack on those using DOH. Many reported error messages saying that Cloudflare's DOH server were practically returning the certificate for Telekom Malaysia's DNS servers.

Even with many new technologies, I ralized that I not as safe and free as I want to be, maybe you too.

It will support PS/2 keyboards and VGA. Intended to be a learning platform for RISC-V coding, it's only got 2k of RAM. (I tried finding a 3rd party source for this announcement but I struck out, so I'm linking the actual olimex webpage. Not affiliated with them, just thought this looked neat.)

A U.S. Navy chief who wanted the internet so she and other enlisted officers could scroll social media, check sports scores and watch movies while deployed had an unauthorized Starlink satellite dish installed on a warship and lied to her commanding officer to keep it secret, according to investigators.

Internet access is restricted while a ship is underway to maintain bandwidth for military operations and to protect against cybersecurity threats.

The Navy quietly relieved Grisel Marrero, a command senior chief of the littoral combat ship USS Manchester, in August or September 2023, and released information on parts of the investigation this week.

cross-posted from: https://slrpnk.net/post/13254852

Theres none so blind as those who don't want to see.

This collaboration with Google underscores the importance of web archiving and expands the reach of the Wayback Machine, making it even easier for users to access and explore archived content.

• https://blog.archive.org/2024/09/11/new-feature-alert-access-archived-webpages-directly-through-google-search/

https://therecord.media/ford-patent-application-in-vehicle-listening-advertising

Looks like Ford just voted to NOT be my choice of car vendor in future.

However getting this patent could be used by Ford to prevent such systems from beeing used by all vendors, but thats veery unlikely i guess.

cross-posted from: https://lemm.ee/post/41907712

cross-posted from: https://lemmy.world/post/19421887

DeArrow is an open source browser extension for crowdsourcing better titles and thumbnails on YouTube. The goal is to make titles accurate and reduce sensationalism. No more arrows, ridiculous faces, and no more clickbait.

"Clickbait" isn't the exception anymore, it's becoming the norm. Many have even started going through their entire backlog, changing old titles and thumbnails to be more attention grabbing and vague.

It's no one's fault. It's a system that creates a race to the bottom.

DeArrow hopes to stop this cycle. It's time to return to a more peaceful experience.