Piyush Malik

#AI is Running Large Scale Experiments on Humans How is it we have to take survey questionnaires through a rigorous review process and drugs have extensive trials that take 10-15 years before they can come to market but technology with far-reaching and unknown impacts like AI can be introduced without any external assessment? Does that strike you as strange? It does me. https://lnkd.in/erbPvD4s

13

5 Comments

📢 Who Invented the Term QoE (Quality of End User Experience)? 📢 Ever wondered who coined the term QoE (Quality of End User Experience) and pioneered solutions to measure it, relating it to QoS (Quality of Service) for the first time nearly 20 years ago? Discover the innovator behind this groundbreaking concept and the impact it has had on the telecommunications industry. Watch the video below to learn more! 🎥👇 #QoE #drdavidsoldani https://lnkd.in/gcvF8Tux

11

1 Comment

Some scholars are inflating their reference counts by sneaking extra citations into article metadata, distorting citation metrics and giving certain authors an unfair advantage. Bruce Schneier explains how this practice undermines the integrity of scientific publishing. #citationmanipulation #scientificpublishing #researchintegrity #openscholarship #academicintegrity #this_post_was_generated_with_ai_assistance #responsibleai https://lnkd.in/eEXJBDJG

1

With outcome of supporting research for future brain-inspired artificial intelligence and sustainable AI; researchers at Sandia National Laboratories plan to use Hala Point for advanced brain-scale computing research. The organization will focus on solving scientific computing problems in device physics, computer architecture, computer science and informatics. Neuromorphic computing is a fundamentally new approach that draws on neuroscience insights that integrate memory and computing with highly granular parallelism to minimize data movement.  Hala Point is the first large-scale neuromorphic system to demonstrate state-of-the-art computational efficiencies on mainstream AI workloads. For more on the news check-out the Business Wire article: https://lnkd.in/gN82wari #IntelAI #SustainableAI #NeuromorphicComputing #AIeverywhere

77

Accelerate your research with Phi-3-Medium's logic-rich features, now available in both short (4K) and long (128K) contexts! Chintan Shah explains how this can benefit your work. #research #AI #logic #contextlength https://lnkd.in/eCPjvaM9

1

There is so much noise in the data & AI space - who should you listen to? Today, I’d like to highlight 5 voices who are sharing true actionable insights and leading us into the future. Peggy tsAI is the CDO at BigID and brings to you cutting edge advice on AI & data management implementations. Elena Alikhachkina, PhD has piloted multi-billion dollar projects for Fortune 500s which allows her to bust myths, call out the hype and share what is working.  She is a my go-to person on all things data products Cindi Howson is the host of one of my favorite podcasts - The Data Chief. If there is a recognition or award in our space, she has probably already won it. Olga Maydanchik is the go-to expert on data quality management, and one of the very few people out there who know how to enable active metadata in an enterprise. Susan Walsh is the classification guru & fixer of dirty data. Her COAT framework will help you avoid costly business mistakes because of data mismanagement. If you want to stay ahead of the curve, follow these 5 genuine, innovative voices in the data & AI space. Do you have other experts in mind? Tag them in the comments below!

64

10 Comments

One of the simpler ways to learn about trends in #genai is to just read the newsletter by Andrew Ng . The second best thing to do is to take the short courses he has launched though DeepLearning.AI

2

In recent years, the field of speech recognition has witnessed significant advancements, largely due to the adoption of Hidden Markov Models (HMMs). These models have become foundational in developing robust and efficient speech recognition systems. This blog post explores the reasons behind the dominance of HMMs in this area, focusing on two key approaches: their mathematical rigor and their training on large corpora of real speech data.

1

1 Comment

ChatGPT went down yesterday- which means it's critical that you know about other tools out there. Let's learn from The Wall Street Journal's bot challenge. Want to follow the top AI news and become a genAI ninja in the process? Go to www.conorgrennan.com to sign up for the free AI Mindset Newsletter.

77

40 Comments

Another demonstration of ScienceCast.org #AI enhancements of arXiv postings - free, simple interface https://lnkd.in/e_gXsF_9

3

1 Comment

Spatial intelligence is making the impossible possible for digital minds. Fei-Fei Li and her lab are at the forefront of embodied AI as described in her TED talk. #DecodingAI®️ #CFiAI https://lnkd.in/eESq-kcS?

4

What happens when two industry veterans and former analysts meet and have a private chat about the marketing hyperboles and product overpromise of tech vendors? What do they say to each other? Welcome to Reality Distortion Fields (RDFs)! A new long form conversation podcast hosted by Elias Khnaser and Alessandro Perilli 🇺🇦. In this first episode, Elias and Alessandro talk about: - Humane AI Pin and MKBHD - Rabbit R1 - what's the job of a reviewer/critic/analyst - how companies grow reality distortion fields and if there's a way to avoid them - tech vendors' responsibility towards customers and corporate incentives - how they typically deal with failures, and the contrarian perspective of Peter Thiel - if we really need yet another device - what's the vision for an AI assistant compared to what vendors are selling today Watch it here or wherever you get your podcasts: https://lnkd.in/g9Pdr7P5

12

2 Comments

There is no innovation without collaboration. Innovation thrives when diverse stakeholders—industry, academia, the public sector, and startups—collaborate. Each brings unique strengths to the table, creating a synergy that propels groundbreaking advancements. Industry: Converts theoretical concepts into practical applications, leveraging resources and market access to scale innovations. Academia: Generates cutting-edge research, ensuring it has practical implications through partnerships with industry. Public Sector: Provides funding, regulatory support, and public-private partnerships, fostering an environment conducive to innovation. Startups: Bring agility and disruptive ideas, challenging norms and exploring new business models. When we collaborate we can facilitate: 1. Resource Sharing: Ensures projects have necessary funding, knowledge, and tools. 2. Diverse Perspectives: Fosters creative problem-solving. 3. Risk Mitigation: Spreads risk, enabling ambitious projects. 4. Accelerated Development: Reduces time from concept to market. 5. Broader Impact: Ensures innovations benefit various societal segments. Collaboration is essential for driving innovation. By leveraging each other’s strengths, these stakeholders create a vibrant ecosystem where groundbreaking ideas flourish, benefiting society as a whole. What are your favourite ecosystem examples? #collaboration #ecosystem

12

"The use of synthetic data has the potential to enhance existing data to allow for more efficient and inclusive practices and policies.   However, we cannot assume synthetic data to be automatically better or even equivalent to data from the physical world.   There are many risks to using synthetic data, including cybersecurity risks, bias propagation, and increasing model error. This document sets out recommendations for the responsible use of synthetic data in AI training…   In constructing synthetic data, the source and nature of values driving the generation process significantly impact its relevance and ethical implications.   The distinction between public and private sector values introduces considerations of transparency, accountability, and the potential influence of profit motives.   Moreover, a rights-based approach emphasizes the importance of respecting individual privacy and ensuring that the synthetic data align with legal and ethical standards.   Cultural values further contribute to shaping the context in which the data are generated, influencing decisions about what aspects of a society or community should be represented and how.   Recognizing and navigating these diverse values is crucial for developing synthetic datasets that not only accurately reflect real-world scenarios but also uphold ethical principles, safeguard privacy, and avoid perpetuating biases or inequalities..."  UN University – Institute for the Advanced Study of Sustainability United Nations https://lnkd.in/dRdsaDej

8

The world of natural language processing is abuzz with talk of large language models (LLMs). But there's a rising star in the AI space that deserves our attention: small language models (SLMs). What exactly are SLMs? Think of them as the focused, nimble cousins of LLMs. With fewer parameters, they offer several advantages: Reduced Computational Cost: SLMs are easier to train and deploy, even on devices with less processing power. Task Specialization: SLMs can be fine-tuned for specific domains, leading to better performance in niche areas. Speed and Efficiency: Their leaner architecture means SLMs often have faster inference times for real-time applications. Enhanced Privacy: SLMs can potentially operate locally with less data transfer, improving data security. Where SLMs Shine Customer Support Chatbots: SLMs excel at answering frequently asked questions and resolving common issues, freeing up human agents. Embedded Applications: Integrate SLMs into IoT devices or mobile apps for tasks like translation or content summarization. Domain-Specific Tasks: SLMs can be tailored for industries like healthcare, finance, or law, where specialized vocabulary matters. Edge Computing: SLMs are ideal for resource-constrained environments where low latency and local operations are crucial. Microsoft's Phi-3 is a testament to the potential of SLMs. As researchers continue to advance SLM development, we can expect even more innovative applications. What are your thoughts on the future of SLMs? #languagemodels #AI #NLP #Microsoft #Phi3

12

1 Comment

Phi-3 small language models and VASA-1 image to video tools aka Having Worked for PwC, Can I Mock Microsoft's Branding Strategy? It has become redundant to say that the news in GenAI just keeps coming faster and faster. Microsoft today announced Phi-3 open-source "small language models". (1) A few days ago, they announced VASA-1, which can turn photos into talking faces. Let's start with Phi-3. Would you like to interact with a GPT-3.5 class chatbot on your Android phone without needing a connection to the Internet? While the move to enterprise solutions and AGI (artificial general intelligence, as well as adjusted gross income) are underway, getting AI into the hands of people with less powerful devices who can't be online all the time is another important move. Why "Phi-3"? Yes, there was a Phi-2, a Phi-1.5, and a Phi-1. It comes from a paper, "Textbooks Are All You Need" (2). It may be a reference to the Greek letter Φ φ, the Golden Ratio, or something else. I couldn't quickly find an explanation. It looks like a partial score of a Phillies, Eagles, or Flyers game. (Phi-3 7, Buf-3 4, sigh.) But Phi does not bother me as much as other MS names. I wrote about my concerns with the name "Deucalion" previously. (3) Now we have their "Vasa-1" AI tool for making lifelike video from a single photograph. (4) The Vasa is the name of a famed Swedish warship. OK, an infamous 17th century Swedish warship, the anchor exhibit of one of my favorite museums. The Vasa is famous for ... wait for it ... sinking almost immediately after leaving dock on its maiden voyage. Hooray! Gust of wind. Down she goes. (5) Raised after some 333 years, incredibly well preserved by the cold and brackish waters, it now sits in a museum at which my namesake Sankt Erik, an icebreaker, is an exhibit as well. I like to begin my presentations with an icebreaker. For those who don't remember PwC's naming history, they received a lot of pushback for: Monday - the tentative name of a spinout of the consulting business during the dot-com/bomb era PwC - personal water craft? Strategy& - the & was an escape sequence; you couldn't search for the name when it first came out PwC - I admire you! Vasa - I appreciate you! (1) https://lnkd.in/gpmd_8ZP, https://lnkd.in/gtCKByKd (2) https://lnkd.in/gQ9QvkRE (3) https://lnkd.in/gcgiwJK9 (4) https://lnkd.in/gJqNKk-z (5) https://lnkd.in/gN4xB-xD)

7

The threat of AI in the hands of bad actors is real. I'm interested in seeing how this maps to the underlying human intent. This chart tells me that, like a kitchen knife, AI is a tool that can be weaponised. As usual, the problem is people, not technology. #pessimism #technology #ai #genai #risk #riskmanagement #riskmitigation #threats #existentialthreats #badactors #chart #humancondition #weapons #people #humanbehaviour #danger #intent #bollox #informationsystems #propaganda #manipulation #profit #monetisation #fraud #scams #parody #satire #cyberattacks #subversion #environment #ecology #energy

1