- AI Emergence
- Posts
- 100x more powerful AI than GPT-4?
100x more powerful AI than GPT-4?
Along with: Who Knew 122 Days Were Enough to Build an AI Giant? Musk Did!
Hey there,
This week gave an astonishing peek into what could be coming in the next few quarters! On one side, the CEO of OpenAI Japan teased a 100x more powerful AI and on the other Elon announced xAI’s Colossus with 100,000 GPUs and a promise to add as many GPUs in the coming weeks!
Both the data points clearly defy the pace of innovation we have seen going live since GPT4 went live - exciting times ahead!
What would be the format? Every week, we will break the newsletter into the following sections:
The Input - All about recent developments in AI
The Tools - Interesting finds and launches
The Algorithm - Resources for learning
The Output - Our reflection
Table of Contents
Tadao Nagasaki, CEO of OpenAI Japan, announced at the KDDI Summit 2024 that GPT-Next, OpenAI's upcoming AI model, will be 100 times more powerful than GPT-4.
This leap in performance is attributed to improvements in architecture rather than an increase in computing resources, and the model will use a smaller version of the mysterious Project Strawberry.
Nagasaki highlighted the rapid growth of ChatGPT, which now has over 200 million active users, making it the fastest software in history to reach this milestone. He also noted the increasing adoption of ChatGPT Enterprise by companies and reaffirmed OpenAI’s commitment to advancing AI technology.(source)
Recently, Musk announced that his team at xAI has brought the Colossus 100,000 H100 training cluster online, and that too in 122 days.
Now the most powerful system in AI training, this system surpasses OpenAI’s 80,000 GPUs. Developed in collaboration with Nvidia, Colossus features state-of-the-art H200 GPUs and promises exceptional energy efficiency. (source)
MiniMax, a Chinese AI startup by Alibaba and Tencent, launched its text-to-video generation model, Video-01, which can create six-second videos in various styles and resolutions.
Yan Junjie, the company's CEO, introduced the platform during their first developer conference in Shanghai.
Video-01 is publicly accessible and adds to the company’s existing AI tools like music generation and speech synthesis. Although the tool competes with similar platforms like Runway AI and OpenAI’s upcoming Sora, concerns around deepfakes and data privacy remain unaddressed.(source)
Amazon is set to launch an enhanced version of Alexa, called "Remarkable Alexa," powered by Anthropic’s Claude AI model.
Initial trials of Alexa’s in-house AI models fell short, leading Amazon to adopt Claude for better performance. This update comes amid scrutiny of Amazon’s $4 billion investment in Anthropic by UK competition regulators.
Despite challenges and delays, Remarkable Alexa, expected in mid-October, will introduce features like daily AI news summaries, a child-friendly chatbot, and conversational shopping tools. Amazon plans to introduce this upgraded Alexa through a subscription model while keeping the Classic Alexa free. (source)
Magic AI has introduced the LTM-2-Mini, a new language model designed for efficient text generation and natural language understanding. This model utilizes a transformer-based architecture to process and generate text with high coherence while minimizing computational costs.
Magic AI is also pioneering ultra-long context models capable of reasoning with up to 100 million tokens, which promises significant advancements in software development. Their new evaluation tool, HashHop, improves the assessment of long-context capabilities by avoiding flaws in traditional methods.
Backed by a $465 million funding boost and strategic partnerships with Google Cloud and NVIDIA, Magic AI is set to enhance its training and inference efficiency. The company also seeks talented engineers and researchers to join their innovative projects. (source)
Bland AI has raised $16 million in Series A funding to advance its platform, which automates enterprise phone communications with realistic AI agents.
It seeks to replace traditional call centers with AI solutions capable of handling customer support, sales, and internal communications more efficiently than human operators.
The platform, featuring voice cloning and multi-language support, has already gained traction with companies like Better.com and Sears.
Despite its innovative approach, Bland AI faces ethical concerns about its agents potentially misleading users into thinking they are human, a practice known as “human-washing.”(source)
OpenAI is actively developing its own AI chips, securing production capacity from TSMC’s advanced 1.6nm A16 process node, set to begin mass production in 2026.
This strategic move involves collaboration with Apple, Broadcom, and Marvell to create custom ASIC chips.
Apple, already integrating OpenAI’s technology into its AI systems, stands to gain from these developments.
OpenAI’s AI model, Sora, which converts text into video, will likely benefit from these advanced chips, enhancing Apple’s competitive edge in the AI market and driving demand in related industries such as optical communications and data transmission. (source)
China's push for autonomous vehicles is sparking job security fears among drivers, but experts believe that the technology is already creating new opportunities in areas like vehicle testing, system monitoring, data annotation, and cybersecurity.
China issued 16,000 test licenses for autonomous vehicles and opened 32,000 kilometers of public test roads. Companies like Baidu's Apollo Go, which operates robotaxis, are expanding rapidly and offering new roles within the autonomous vehicle ecosystem.
Although the transition may be slow and phased, the technology is expected to create more jobs in the long run, offsetting some of the job losses. (source)
1X has introduced NEO Beta, a bipedal humanoid prototype designed for home use, marking a significant milestone as the company transitions from conceptual development to integrating humanoid robots into consumer households.
NEO Beta features bio-inspired designs that ensure safe interaction with humans. It will undergo testing in selected homes as part of its research and development phase.
With this launch, 1X aims to create an abundant supply of physical labor through intelligent humanoids capable of performing a wide range of tasks. (source)
Tool: Toolsaday
We spend about 30% of our day either composing emails or replying to them. Often, we rely on chatbots for grammar assistance but frequently find ourselves explaining multiple times how we want the email to be written. Toolsaday simplifies this process. You only need to specify the topic, desired length, and tone, and the tool will generate the email for you.
How to access:
Log in to the Toolsaday.
Specify the purpose of the email.
Choose the type of subject line you need.
Select the desired length and tone.
Generate the email instantly.
This tool will save you significant time by automating email writing and ensuring correct grammar and tone, allowing you to focus on more important tasks.
In the recent episode of Leading with Data, I had an interesting conversation with Gaurav Agarwal, the Founder and CEO of RagaAI. He shared his views on building high-performing teams in the AI industry and how it plays a crucial role in preventing failures and driving innovation. We also explored his journey in AI, the inspiration behind RagaAI, and his advice for aspiring professionals in this fast-paced field.
If you want to leverage Python for AI without necessarily becoming a software developer or AI engineer. This course by deeplearning.ai is about teaching Python programming with a focus on building practical AI applications. The course offers personalized support through an AI chatbot, making it accessible to anyone looking to enhance their programming skills in today's data-driven world.
I got a chance to hear a recent podcast of Lex Fridman with Pieter Levels and it left me inspired about what each of us can do in coming years.
That’s it for the week - I look forward to the Apple Event early next week and I hope Apple announces a clear roadmap for its Apple Intelligence.
What are you looking forward to this week?
How do you rate this issue of AI Emergence?Would love to hear your thoughts |
Reply