The artificial intelligence (AI) realm witnesses rapid developments around the world, reshaping the world of digital transformation and accelerating the pace of AI adoption. In this weekly newsletter, FollowICT is highlighting the most important news and advancements in AI around the globe.
Chinese AI S tartup Manus Secures $500 MLN in Funding
Chinese AI startup Manus, which is building tools related to AI agents, has secured $75 million in a funding round led by Benchmark at a valuation of about $500 million, according to Bloomberg. Bloomberg reported, citing people familiar with the matter, that the company will use the funds to expand into new markets, including the United States, Japan, and the Middle East. The Bloomberg report indicates that the new round has quintupled Manus’ valuation, which previously raised over $10 million from backers including Tencent and HSG (formerly known as Sequoia China).
Perplexity Browser To Track Users Online Attitude to Sell “Hyper-personalized” Ads
CEO Aravind Srinivas said on the TBPN podcast this week that one of the reasons Perplexity is developing its own browser is to collect data on everything users do outside of its own app. This is so it can sell premium ads. “That’s one of the other reasons why we’re building a browser, which is we want to get data even outside the app to understand you better. Because some of the queries that users ask these systems are just work-related, not personal,” Srinivas said. Work-related queries won’t help the AI company build a detailed enough profile. “On the other hand, what are the things you buy; what are the hotels you’re going to; what are the restaurants you’re going to; what do you spend your time browsing, tells us a lot about you,” he added.
Dropbox adds new features to Its AI-powered Dash Search Tool
Dropbox released on Thursday an update to its AI-powered search tool, Dash, which first launched in 2023. The company is adding AI “understanding” of different content types to Dash, letting users search audio, video, and images in addition to text.
The company is also adding people search to let users look for someone who worked on a specific project or is an expert in a certain area. Last year, Dropbox unveiled Dash for Business to let companies use AI search. This year, the company is improving Dash’s tools for organizations by adding support for IT admins to exclude certain sensitive documents from search results.
Dropbox already offers functionality to summarize documents using AI. With this new release, the company is introducing new writing tools that leverage summaries from different data sources to generate new documents and presentations. The company explained that these tools, available in Dash, can pull information from email, meeting notes, and existing documents to create project plans, memos, or briefs. The core idea is that users won’t have to switch between apps to read some information and add it to a document.
OpenAI Launches “Lightweight” Version of ChatGPT Deep Search tool
ChatGPT OpenAI announced it will offer a new “lightweight” version of its ChatGPT Deep Search tool, which trawls the web to compile research reports on a topic, to ChatGPT Plus, Team, and Pro users. The new lightweight Deep Search, which will also be available to free ChatGPT users starting today, runs on a version of OpenAI’s o4-mini model, according to OpenAI. While it’s not as capable as the “full” Deep Search, OpenAI claims its service is less expensive, enabling the company to increase usage limits. “Responses are typically shorter while maintaining the depth and quality you’ve come to expect,” OpenAI said in a series of posts on X. “Once you hit the limits for the original Deep Search, queries will automatically fall back to the lightweight version.”
Elon Musk’s xAI Holdings in Talks for 2nd-largest Funding Round Ever
Elon Musk’s xAI Holdings is in talks to raise a new funding round of $20 billion, which could value the AI and social media company at over $120 billion, according to a new report from Bloomberg, which noted that the talks are still in early stages. If successful, the deal would be the second-largest funding round for startups ever, behind OpenAI’s $40 billion round last month.
The funding could help ease the heavy debt burden at X, which costs the company a whopping $200 million per month in servicing fees, according to Bloomberg’s sources, with annual interest expenses surpassing $1.3 billion by the end of last year.
AI Code-scanning Tool Maker Endor Labs nabs $93 MLN
American AI code-scanning tool maker Endor Labs, which is building tools to scan AI-generated code for security vulnerabilities, has raised a $93 million Series B, led by DFJ Growth with participation from Salesforce Ventures and its existing investors. The company indicated that this isn’t just funding, it’s fuel for what they’ve been building: the most comprehensive AppSec platform for the AI era. They added, “As AI reshapes how code is written, we’ve been heads down on how to secure it with: – Code intelligence from over 4.5M OSS projects – AI agents that think like developers, architects, and security engineers – Real remediation, not just scanning Today, we secure over 5M applications, run over 1M scans per customer weekly, and have grown ARR 30x with 166% net revenue retention, making Endor Labs the fastest growing AppSec company ever.”
Adobe Launches New Firefly Image-generating Models
Adobe Adobe has launched the latest iteration of its Firefly AI image-generating models, a vector generation model, and a redesigned web app that houses all of its AI models, as well as some from competitors. A mobile app for Firefly is also in the works. Adobe says its new Firefly Image Model 4 improves upon previous versions in terms of quality, speed, control over the structure and style of outputs, camera angles, and zoom. It can produce images with up to 2K resolution.
An improved and more efficient version of this model, called Image Model 4 Ultra, is also available, which can render complex scenes consisting of small structures and lots of detail. The company is also making its Firefly Video model, which launched in limited beta last year, available to everyone. This model lets users create videos from text or an image, use camera angles, specify start and end frames to control shots, create atmosphere elements, and customize motion design elements. The model can generate videos from text with up to 1080p resolution.
xAI’s Grok Chatbot Can Now See the world Around it
xAI’s Grok chatbot can now answer questions about what appears in your smartphone camera’s field of view, similar to real-time vision features available in Google’s Gemini and OpenAI’s ChatGPT. xAI announced Tuesday the launch of Grok Vision, which lets users point their phones at things like products, signs, and documents and ask questions about them.
Grok Vision is available via the Grok app for iOS, but it’s not yet available on the Grok app for Android. Other new features launched for the Grok app include multilingual audio and instant search using voice mode. Grok users on Android can take advantage of these features, but only if they subscribe to xAI’s $30-per-month SuperGrok tier.
AI Unveils AvatarFX, an AI Video Model for Creating Lifelike Chatbots
AI, a leading platform for chat and role-playing with AI-generated characters, has unveiled its highly anticipated video generation model, AvatarFX. Available in a closed beta, the model animates the platform’s characters in diverse styles and voices, from human-like personas to 2D animal cartoon figures. AvatarFX distinguishes itself from competitors, such as OpenAI’s Sora, by not being solely a text-to-video generator. Users can also create videos from pre-existing images, allowing them to animate photos of real people. –
Baidu Touts New Kunlun Chip Capacity, launches Two AI Models
Chinese search engine giant Baidu said on Friday its in-house developed third-generation Kunlun P800 series of 30,000 chips was now up and running, allowing for the support of training DeepSeek-like models, its chief executive said. Robin Li made the announcement during the company’s annual Baidu World 2024 developer conference, where Baidu, China’s top search engine firm, gave updates on its AI efforts, according to Reuters.
Li said the P800 series of chips could support the training of DeepSeek-like models with hundreds of billions of parameters, and that developers did not need to worry about a shortage of computing power. Separately, Li unveiled Baidu’s latest large language model, Ernie 4.5 Turbo, saying it matched the top models in the industry in several performance tests, showcasing capabilities ranging from programming to language comprehension. The company also launched on Friday a new reasoning model called Ernie X1 Turbo.
Both of Baidu’s new models are multimodal. Baidu said it would integrate AI capabilities into its applications, with Li saying: “There are many (AI) models, but applications are what rule the world. Application is king. Without applications, models and chips are worthless.”
China’s Xi calls for Self-reliance in AI Development amid US rivalry
Chinese President Xi Jinping pledged “self-reliance and self-strengthening” in developing artificial intelligence in China, state media reported on Saturday, as the country vies with the United States for leadership in AI, a key strategic area. Speaking during a study session of the Politburo meeting on Friday, Xi said China must leverage its “new whole-nation system” to advance AI development. “We must recognize the gaps and redouble our efforts to comprehensively enhance the technological innovation, industrial development and security capabilities underpinned by AI,” Xi said, according to the official Xinhua news agency. Xi indicated that policy support would be provided in areas such as government procurement, intellectual property rights, research and talent development.
Understand AI
Large Language Model (LLM)
A Large Language Model is a type of artificial intelligence model that has been trained on massive amounts of text, enabling it to understand human language and generate new text in a way that closely resembles human writing. The model reads millions or billions of words from the internet, books, articles, and learns to accurately predict the next word in a sentence. Through this training, it becomes capable of writing articles, answering questions, composing stories, and even holding full conversations.
Famous examples:
ChatGPT (from OpenAI)
Claude (from Anthropic)
Gemini (from Google DeepMind)
LLMs are the heart behind AI applications that deal with text and speech. Today, these models are used in education, customer service, media, and even programming.
