- Dev Notes
- Posts
- Google’s NEW AI Model “Gemini” > GPT-4
Google’s NEW AI Model “Gemini” > GPT-4
Good Morning! Google unveils new multimodal AI model Gemini, claiming it outperforms OpenAI's GPT-4 in benchmarks. JetBrains incorporates AI capabilities into its IDEs to assist developers with coding tasks through an intelligent assistant. Senator Wyden's letter exposes governments compelling Apple and Google to secretly provide data on user push notifications for surveillance.
Google’s NEW AI Model “Gemini” > GPT-4
Google has announced its most advanced AI model yet, called Gemini. As a multimodal model, Gemini can understand and process text, images, audio, video and even code. It comes in three sizes - Gemini Nano, Gemini Pro and Gemini Ultra - to work across devices from phones to data centers.
Google is positioning Gemini as superior to OpenAI's popular GPT-4 model for certain tasks. Benchmark tests show Gemini Ultra outperforming GPT-4 in 30 out of 32 common AI research tests. However, some experts argue Google may be exaggerating or misrepresenting Gemini's capabilities.
For now, Gemini is only available in a limited capacity. The Gemini Pro version powers Google's Bard chatbot, while Gemini Nano enables faster AI-powered suggestions in Pixel phones. Google plans a broader launch of Gemini in 2024, integrating it across Search, Maps and more.
Key Takeaways:
Gemini is Google's newest AI able to understand text, images, audio and more
Comes in three sizes to work on phones, computers or data centers
Google claims benchmark tests show Gemini outperforming GPT-4
Critics argue Google may be exaggerating Gemini's abilities
Currently limited access; broader launch with more Google products in 2024
Read More Here
JetBrains adds AI Assistant to all of its IDEs
JetBrains has announced the addition of an AI Assistant across all of its IDEs to help developers be more productive. Available now in the latest versions of products like IntelliJ IDEA, PyCharm, and WebStorm, the AI Assistant aims to save developers time on common coding tasks.
"Empowering developers with AI-backed features supports the long-standing mission of JetBrains: to improve your experience by automating common and repetitive tasks, enabling you to stay in the flow and focus on the big picture," said Svetlana Zemlyanskaya, team lead for AI Assistant at JetBrains.
The AI Assistant can answer developer questions about their code, provide intelligent code completion suggestions, generate documentation, and more. It's powered by a combination of OpenAI models and JetBrains' own proprietary models, with plans to support third-party AI models in the future.
JetBrains says the deep integration with IDEs gives the AI Assistant strong context about developer projects to make highly relevant recommendations. At the same time, only necessary data is shared and isn't used for training generative AI models.
AI Chat - Answers coding questions by considering project context
Code Completion - Predicts next lines of code based on context
Refactoring - Suggests and explains potential code improvements
Documentation - Generates documentation for functions/classes
Commit Messages - Summarizes code changes to simplify commit notes
The AI Assistant is available now to existing JetBrains IDE customers under a separate subscription for the JetBrains AI service. Over time, it aims to automate repetitive programming tasks to help developers focus on big-picture thinking.
Read More Here
Apple and Google confirm governments spy on users through push notifications
A concerning new privacy issue has come to light - governments are able to spy on smartphone users by accessing data from push notifications. This was revealed in a letter written by US Senator Ron Wyden to the Department of Justice.
Push notifications, which apps use to send alerts and messages directly to users, pass through servers operated by Apple and Google. This gives those companies access to metadata about notifications sent to devices. Senator Wyden's letter indicates that governments have been secretly compelling Apple and Google to hand over this type of data to surveil users. The information gained could show a user's interactions with particular apps, provide the full text of notifications, and even reveal some unencrypted content. Essentially, governments have a backdoor through mandatory centralization of push notifications that allows spying on users without their knowledge or consent.
In his letter, Senator Wyden urged the Justice Department to allow Apple and Google to be transparent about government demands for push notification data. In response, Apple stated that they had previously been prohibited by the federal government from sharing details publicly. Now that the surveillance method is known, Apple says it will start documenting government data requests of this nature. Google also acknowledged the senator's concerns, saying the company is committed to keeping users informed about government requests. Google already includes some aggregate data on these types of demands in its regular transparency reports.
This news confirms fears about government overreach and surreptitious erosion of civil liberties. Push notification spying exploits centralized elements of mobile ecosystems in ways average users would never expect. Although the scale is still unclear, the surveillance capability clearly has enormous potential for monitoring and tracking device owners without judicial oversight. For individuals and digital privacy advocates alike this revelation underscores deepening risks in an increasingly connected world.
Read More Here
New Firmware Vulnerability Puts Most PCs at Risk
A dangerous new vulnerability has been discovered that could allow hackers to infect the firmware of most Windows and Linux PCs, according to security researchers. The attack, dubbed "LogoFAIL", targets the early boot process before operating systems load.
The vulnerability was uncovered by security firm Binarly after nearly a year of research. It exploits image parsing bugs in the firmware interface used to boot devices. By replacing legitimate manufacturer logos displayed early in the process with malicious lookalikes, attackers can run malware.
Once executed, LogoFAIL grants full control over a device's memory and storage. It can then infect the hard drive with stealthy malware before the operating system boots. This allows the attack to bypass firmware defenses like Secure Boot that are meant to prevent bootkits.
The vulnerability impacts hundreds of devices from major manufacturers like Dell, HP, and Lenovo. CPU vendors like Intel and AMD are also affected. Even systems running Secure Boot, Intel Boot Guard, and other boot protections are susceptible.
LogoFAIL highlights risks in the unified extensible firmware interface (UEFI) that has replaced BIOS in modern computers. While offering more flexibility, UEFI contains complex components from multiple vendors that expand the attack surface. The ubiquitous logo images displayed while booting proved to be a stealthy attack vector hiding in plain sight.
The wide range of affected devices makes LogoFAIL a critical threat. UEFI has clearly emerged as a security trouble spot in dire need of scrutiny from both vendors and the security community. Companies are releasing advisories and patches, but the challenging firmware update process could hamper mitigation efforts. For now, all Windows and Linux users should check with their device manufacturer and apply updates as soon as available.
Read More Here
Youtube Spotlight
The Making of VIM: The Most Beloved Text Editor of All Time
Vim, originally created in 1988 as a VI port for the Amiga computer, has evolved into one of the most beloved and popular text editors worldwide. Its development includes a rich history from the creation of the Ed text editor in 1969, the evolution of VI, and the introduction of Stevie in 1987. Through open-source principles and community-driven enhancement, Vim has become more than just a text editor, but an ecosystem with a unique approach to coding, captivating programmers from around the world.
Was this forwarded to you? Sign Up Here
Reply