In July’s monthly digest, we’ll examine the ongoing search engine showdown between Google, Microsoft Bing, and OpenAI to determine who leads the pack. We’ll also explore the latest advancements in AI models and key developments in the AI landscape alongside updates from the world of robotics. There is a lot to discuss. Ready to get started?
Search engine battles: who runs the world?
Google is determined to stay ahead of the competition and tries its best not to lose the first place. Recently, the tech giant announced updates to its Gemini chatbot to make it faster and more accessible. The new version, Gemini 1.5 Flash, is now available on the web and mobile in 40 languages across 230 countries. Gemini 1.5 Flash boasts improvements in speed and performance, especially in reasoning and understanding images. It is not only better, but it might also be cheaper for Google to operate. Google also expands Gemini’s context window, handling more extended and complex inputs.
Not to be outdone by Google, OpenAI has unveiled its new tool, SearchGPT. Launching as a prototype, SearchGPT uses real-time web data to provide quick answers and relevant links based on user queries. Similar in interface to ChatGPT, SearchGPT offers location-based results and integrates OpenAI’s GPT-3.5 and GPT-4 models. It aims to improve search efficiency while addressing AI content issues by citing and linking to original publishers. Initially available to select users, SearchGPT represents OpenAI’s push to innovate in search technology, promising a more responsible approach to AI-generated content.
Without wanting to be left behind by competitors, Microsoft has introduced its version of AI-powered search: Bing Generative Search. Currently available to only a tiny percentage of users, this new feature combines various generative AI models to gather and summarize information from across the web in response to search queries. Users can choose to see traditional search results instead of the AI-generated summary. They plan to expand its availability gradually while working to improve its performance. This launch follows Microsoft’s previous efforts to integrate AI into Bing. Microsoft aims to deliver more accurate search results while navigating the challenges seen with other AI search tools, such as Google’s AI Overviews and others that have faced issues with accuracy and attribution.
AI
AI models showdown
Meta has launched its largest open-source AI model, Llama 3.1 405B, featuring 405 billion parameters. This model, trained on 16,000 Nvidia H100 GPUs, is notable for its size and performance, though it’s not the largest open-source model currently available. Llama 3.1 405B excels in tasks like coding and summarizing text in eight languages but remains text-only, without multimodal capabilities like image or speech recognition. Meta is also experimenting with multimodal models and using refined training data and synthetic data for this release. The new model has a larger context window, allowing it to handle more extensive text inputs. Meta aims to expand its AI ecosystem by providing developers with various tools and licensing options.
In a related development, OpenAI has introduced GPT-4o mini, its latest small AI model, which is both cheaper and faster than its previous models. Available to developers and consumers via the ChatGPT app starting today, it will be accessible to enterprise users next week. It supports text and vision now, with plans to add video and audio capabilities. The model boasts a large context window of 128,000 tokens and is noted for its speed, being over twice as fast as GPT-4o and GPT-3.5 Turbo.
Meanwhile, Mistral has unveiled its new flagship model, Large 2, positioning it as a competitor to Meta’s Llama 3.1 405B and OpenAI’s latest models. With 123 billion parameters, Large 2 surpasses Llama 3.1 in code generation and math performance while being more cost-effective. The model features a large 128,000-token context window and enhanced multilingual support. Mistral aims to address hallucination issues by improving response accuracy and is available on major cloud platforms and Mistral’s own services.
OpenAI is accused of illegally restrictive NDAs
Again OpenAI is at the center of scandal. Whistleblowers have accused OpenAI of enforcing “illegally restrictive” non-disclosure agreements (NDAs) that could prevent employees from reporting harmful behavior to government agencies. The whistleblowers’ lawyers have contacted SEC Chair Gary Gensler, claiming that OpenAI’s NDAs force individuals to seek prior consent before sharing information with federal authorities, potentially discouraging whistleblowing. This controversy comes amid notable departures from OpenAI, including co-founder Ilya Sutskever. For further details, see today’s TechCrunch Minute.
Wiz turns down Google’s $23B acquisition offer
Google wanted to purchase cybersecurity startup Wiz for $23 billion, significantly higher than its last private valuation of $12 billion. Despite the attractive offer, Wiz’s management, led by CEO Assaf Rappaport, chose to remain independent. Given their strong team, Rappaport conveyed to 1,200 employees that rejecting such a substantial offer was challenging but necessary. Wiz, founded in 2020 by former Israeli military officers, previously co-founded Adallom, which Microsoft acquired for $320 million.
Latest from robotics
Elon Musk revealed that Tesla plans to start selling its humanoid robot, Optimus, in 2026. This news comes as part of Tesla’s upcoming earnings report, where Musk highlighted Optimus’s current capabilities. Optimus is already performing tasks autonomously at Tesla’s facilities. For instance, it’s been handling batteries and other functions within the company. Musk’s long-term vision includes a significant market for humanoid robots, estimating that demand could reach over 20 billion units globally.
However, Tesla isn’t alone in this rapidly evolving field. Several other companies are making strides in humanoid robotics. For example, this February, Figure, a robotics company specializing in developing humanoid robots, secured a substantial $675 million in funding from major investors like Microsoft, OpenAI, Amazon, Nvidia, and Intel Capital, valuing the company at $2.6 billion. Earlier this year, Figure announced that its robot, the Figure 01, would be piloted at a BMW factory in South Carolina. In 2023, Norwegian startup 1X raised $23.5 million, backed by high-profile investors, including Tiger Global and OpenAI, and later secured $100 million in Series B funding. The company has been showcasing its voice-responsive, wheeled robot, Eve, which performs tasks like cleaning.
Robots can be incredibly beneficial, and numerous startups, such as Ory Laboratory, led by CEO Kentaro Yoshifuji, are leading the way in harnessing their potential. Researchers at Yale University’s Faboratory have developed soft robots with innovative abilities, such as self-amputation and merging with other robots. These soft robots can detach a limb to escape obstacles and fuse to bridge gaps using heat-activated, flexible joints made from a special thermoplastic foam and sticky polymer.
Meanwhile, Waymo has started testing a new robotaxi in San Francisco, which Chinese automaker Zeekr developed. This new model features Waymo’s latest sixth-generation hardware and is designed to be more cost-effective and adaptable to various weather conditions. Currently, only a few of these vehicles are on the road and are not yet autonomous. Waymo’s testing includes extreme heat and cold to ensure reliability. The Zeekr prototypes, assembled in China but equipped with Waymo’s technology in the U.S., are part of Waymo’s strategy to advance autonomous vehicle development amid regulatory discussions on connected vehicles.
Worldwide tech outage
In July 2024, a major global tech outage was triggered by a faulty software update from the cybersecurity firm CrowdStrike, which affected Microsoft Windows systems. This disruption impacted approximately 8.5 million devices, or less than 1% of global Windows machines, leading to significant problems across various industries, including airlines, banking, and healthcare.
Flights were grounded, financial services were interrupted, and many public services faced major issues. However, the problem was not a result of a cyberattack but rather a defect in the update, which exposed vulnerabilities in interconnected digital infrastructures. While systems have largely been restored, the incident underscored the critical need for robust safeguards in technology updates.
