91% Leaders Expect Productivity Gains From AI: Deloitte Survey
Plus: TrustLLM measuring the Truth in LLMs, Tencent launched PhotoMaker.
Hello Engineering Leaders and AI Enthusiasts!
Welcome to the 189th edition of The AI Edge newsletter. This edition brings you: Deloitte Survey Finds 91% of Leaders Expect Productivity Gains From AI.
And a huge shoutout to our incredible readers. We appreciate you😊
In today’s edition:
📊 91% leaders expect productivity gains from AI: Deloitte survey
🔍TrustLLM measuring the Trustworthiness in LLMs
🎨 Tencent launched a new text-to-image method
💡 Knowledge Nugget: The Rise of Sparse Mixtures of Experts: Switch Transformers by
Let’s go!
91% leaders expect productivity gains from AI: Deloitte survey
Deloitte has released a new report on GenAI, highlighting concerns among business leaders about its societal impact and the availability of tech talent. They surveyed 2,835 respondents across 6 industries and 16 countries, finding that 61% are enthusiastic, but 30% remain unsure.
56% of companies focus on efficiency, and 29% on productivity rather than innovation and growth. Technical talent was identified as the main barrier to AI adoption, followed by regulatory compliance and governance issues.
Why does this matter?
The report connects to real-world scenarios like job displacement, the digital divide, issues around data privacy, and AI bias that have arisen with new technologies. Understanding stakeholder perspectives provides insights to help shape policies and practices around generative AI as it continues maturing.
TrustLLM measuring the Trustworthiness in LLMs
TrustLLM is a comprehensive trustworthiness study in LLMs like ChatGPT. The paper proposes principles for trustworthy LLMs and establishes a benchmark across dimensions like truthfulness, safety, fairness, and privacy. The study evaluates 16 mainstream LLMs and finds that trustworthiness and utility are positively related.
Proprietary LLMs generally outperform open-source ones, but some open-source models come close. Some LLMs may prioritize trustworthiness to the point of compromising utility. Transparency in the models and the technologies used for trustworthiness is important for analyzing their effectiveness.
Why does this matter?
TrustLLM provides insights into the trustworthiness of LLMs that impact the findings and help identify which LLMs may be more reliable and safe for end users, guiding adoption. Lack of transparency remains an issue. Assessing trustworthiness helps ensure LLMs benefit society responsibly. Ongoing analysis as models evolve is important to maintain accountability and identification of risks.
Tencent launched a new text-to-image method
Tencent launched PhotoMaker, a personalized text-to-image generation method. It efficiently creates realistic human photos based on given text prompts. It uses a stacked ID embedding to preserve identity information and allows for flexible text control. The authors propose an ID-oriented data construction pipeline to assemble the training data.
PhotoMaker outperforms test-time fine-tuning methods in preserving identity while providing faster generation, high-quality results, strong generalization, and a wide range of applications.
Why does this matter?
Provides an efficient way to generate customizable HQ profile photos from text prompts. Useful for social media and gaming. Connects with real-world needs like easily creating personalized avatars and profile images. The ability to flexibly generate realistic photos while maintaining identity has many applications in social platforms, gaming, the metaverse, and beyond.
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: The Rise of Sparse Mixtures of Experts: Switch Transformers
Sparse Mixtures of Experts (MoE) is a technology that allows for the scaling up of models with a computational complexity of O(1). The Switch Transformer is an architecture that demonstrated the impressive scaling properties of sparse MoE, achieving a 7X speed-up in training.
In MoE, the output is modeled using a linear combination of experts controlled by a gate. In soft routing, the output is combined from all experts, while in hard routing, only the most suitable expert is used. Hard routing allows for computational cost savings without sacrificing too much modeling accuracy.
This thoughtful article by
👏👏👏Why does this matter?
MoE holds significant importance in the realm of scaling up models with remarkable computational efficiency. It achieves a noteworthy 7X speed-up in training and addresses the growing demand for scalable models, allowing for efficient training of deep learning architectures without compromising modeling accuracy.
What Else Is Happening❗
🔧 Apple is reportedly shutting its 121-person San Diego AI team in reorganization
The Data Operations Annotations team will be relocated to Austin to merge with the Texas portion. Employees have until the end of February to decide if they will relocate, and if they choose not to, they will be terminated on April 26. The San Diego team is responsible for improving Siri by listening to queries and ensuring accurate responses. (Link)
💼 Microsoft is launching Copilot Pro
A $20 monthly subscription brings AI features to Office apps like Word, Excel, and PowerPoint. It includes generating PowerPoint slide decks, rephrasing paragraphs, summarizing documents, replying to emails, and analyzing data. It offers access to the latest OpenAI models and improvements to image creation. (Link)
🖋️ AI can mimic a person's Handwriting style
Researchers at Abu Dhabi's Mohamed bin Zayed Uni of AI have developed AI technology that can mimic a person's handwriting style based on a few paragraphs of written material. The neural network uses a transformer model to learn context and meaning in sequential data. The US Patent and Trademark Office granted the technology a patent. (Link)
🌐 OpenAI is launching new AI tools to combat misinformation ahead of elections
The company will attribute information from ChatGPT and help users determine if an image was created by its AI software. OpenAI will encode images produced by its Dall-E 3 image-generator tool with provenance information, allowing voters to understand better if images they see online are AI-generated. They will also release an image-detection tool to determine if an image was generated by Dall-E. (Link)
🔋 Microsoft Researchers used AI to design a battery that uses 70% less lithium
Lithium batteries are used in many everyday devices and electric vehicles, but lithium is expensive, and mining it damages the environment. Finding a replacement for lithium is costly and time-consuming, but using AI, the researchers developed a battery that uses less lithium in months. (Link)
New to the newsletter?
The AI Edge keeps engineering leaders & AI enthusiasts like you on the cutting edge of AI. From ML to ChatGPT to generative AI and LLMs, We break down the latest AI developments and how you can apply them in your work.
Thanks for reading, and see you tomorrow. 😊
Thanks for the kind mention, and keep up the great work!