This New Technique Accelerates LLMs By 300x 😎💯
Plus: 'Screenshot-to-Code' builds entire website, Microsoft Research says Hallucination is necessary in LLMs.
Hello Engineering Leaders and AI Enthusiasts!
Welcome to the 155th edition of The AI Edge newsletter. This edition brings you: A new technique “UltraFastBERT”, that accelerates LLMs by 300 times.
And a huge shoutout to our incredible readers. We appreciate you😊
In today’s edition:
😎 This new technique accelerates LLMs by 300x
🌐AI tool 'Screenshot-to-Code' generates entire code
🤖 Microsoft Research explains why Hallucination is necessary in LLMs!
📚 Knowledge Nugget: “Math is hard” — if you are an LLM – and why that matters by
Let’s go!
This new technique accelerates LLMs by 300x
Researchers at ETH Zurich have developed a new technique UltraFastBERT, a language model that uses only 0.3% of its neurons during inference while maintaining performance. It can accelerate language models by 300 times. And by introducing "fast feedforward" layers (FFF) that use conditional matrix multiplication (CMM) instead of dense matrix multiplications (DMM), the researchers were able to significantly reduce the computational load of neural networks.
They validated their technique with FastBERT, a modified version of Google's BERT model, and achieved impressive results on various language tasks. The researchers believe that incorporating fast feedforward networks into large language models like GPT-3 could lead to even greater acceleration.
Read the Paper here.
Why does this matter?
This work demonstrates the potential for exponentially faster language modeling with selective neuron engagement. This breakthrough could help the analysis of vast volumes of textual data for research purposes and expedited language translations.
AI tool 'Screenshot-to-Code' generates entire code
GitHub user abi has created a tool called "screenshot-to-code" that allows users to convert a screenshot into clean HTML/Tailwind CSS code. The tool utilizes GPT-4 Vision to generate the code and DALL-E 3 to generate visually similar images. Users can also input a URL to clone a live website.
All you want to do is: Upload any screenshot of a website and watch AI build the entire code. It will improve the generated code by comparing it against the screenshot repeatedly.
Why does this matter?
By simplifying the process of code generation from images and live web pages, this tool empowers developers to effortlessly recreate designs. This is a remarkable feat in AI, as this tool can help a more intuitive and efficient approach to web development.
Microsoft Research explains why Hallucination is necessary in LLMs!
Microsoft Research + 4 others have explored that there is a statistical reason behind these hallucinations, unrelated to the model architecture or data quality. For arbitrary facts that cannot be verified from the training data, hallucination is necessary for language models that satisfy a statistical calibration condition.
However, the analysis suggests that pretraining does not lead to hallucinations on facts that appear more than once in the training data or on systematic facts. Different architectures and learning algorithms may help mitigate these types of hallucinations.
Why does this matter?
This research is crucial in shedding light on hallucinations. It highlights some unverifiable facts beyond the training data. Also, these hallucinations might be necessary for language models to meet statistical calibration conditions.
Enjoying the daily updates?
Refer your pals to subscribe to our daily newsletter and get exclusive access to 400+ game-changing AI tools.
When you use the referral link above or the “Share” button on any post, you'll get the credit for any new subscribers. All you need to do is send the link via text or email or share it on social media with friends.
Knowledge Nugget: “Math is hard” — if you are an LLM – and why that matters
This article by
discusses the limitations of language models when it comes to solving mathematical problems, particularly multiplication. The author highlights a paper that claims LLMs can solve mathematical problems without a calculator but argues that the results are not as impressive as they seem.
The author compares the performance of LLMs to that of a calculator and concludes that LLMs struggle to generalize and truly understand multiplication. The article suggests that a hybrid approach may be more effective but emphasizes the need for further improvement in LLMs' mathematical abilities.
Why does this matter?
This article delves into the capabilities of language models, especially in solving math problems like multiplication. It underscores the common perception that these models can handle math without calculators but challenges this belief. Comparing their performance to calculators reveals that language models struggle with generalization and a deep understanding of multiplication. It hints at the need to improve how language models learn and process mathematical information.
What Else Is Happening❗
👥 US, Britain, & other countries signed an agreement to ensure AI systems are "secure by design"
The agreement is non-binding, representing a significant step in prioritizing the safety and security of AI systems. The guidelines address concerns about hackers hijacking AI technology and suggest security testing before releasing models. (Link)
💰 Elon Musk's brain implant startup raised an additional $43 Million
Neuralink brought its total funding to $323 million. The company, which is developing implantable chips that can read brain waves, has attracted 32 investors, including Peter Thiel's Founders Fund. (Link)
⏳ NVIDIA delayed the launch of its new China AI chip
Delayed chip H20, designed to comply with US export rules. The delay could complicate Nvidia's efforts to maintain market share in China against local rivals like Huawei. The company had been expected to launch the new chips on 16 November, but server integration issues have caused the delay. (Link)
🤝 Eviden partners with Microsoft to help clients transition to the cloud and utilize Azure OpenAI Service
Eviden will use its expertise in ML and AI to develop joint solutions and expand its AI-driven industry solutions. Their Gen AI Acceleration Program helps organizations leverage AI with complete trust, offering consultancy on Azure and major data platforms. (Link)
👧 A Spanish agency created its own AI Influencer, and she is making upto $11k in a month
A Spanish modeling agency created the country's first female AI influencer, They decided to design her (López) after having trouble working with real models and influencers. (Link)
That's all for now!
If you are new to The AI Edge newsletter, subscribe to get daily AI updates and news directly sent to your inbox for free!
Thanks for reading, and see you tomorrow. 😊