Enterprise AI Models, EU's Blockchain Ambition, China's Metaverse Masterplan , Swiss Wholesale CBDC and much more!
Hi Everyone,
Welcome to QX Snapshots - a weekly recap of the key news on emerging technologies. In this newsletter, you will get a "digest" of latest info on AI, Quantum Technology, Industrial Metaverse and Enterprise Blockchain.
Hope it brings you value :)
[Quantum Technology] No Quantum Leap Yet But Breakthrough May Pave Way for Reliable Quantum Computers. Microsoft researchers claim to have made a significant advance towards developing reliable quantum computers, an area currently limited by high error rates. The team has devised a new way to represent a logical qubit, the quantum version of the classical bit, using hardware stability and a phase of matter characterized by Majorana zero modes. The use of such matter could assist in producing quantum supercomputers with lower error rates. Despite these promising results, experts caution that it is still early in the field, with various methods being explored for different use cases. Currently, the largest quantum computers have only a few hundred noisy physical qubits, but a commercially viable quantum computer would need millions of near-perfect qubits. Despite the hurdles, advancements in quantum computing have been likened to a "moon shot," indicating their potential to dramatically impact technology.
[AI] Reka’s custom AI models for the enterprise. Reka, a new venture from researchers from DeepMind, Google, Baidu, and Meta researchers, has launched with $58 million in funding, seeking to make large language models (LLMs) like GPT-4 more accessible and practical for specific use cases. The company's first product, Yasa, is a multimodal AI "assistant" trained to understand text, images, videos, and tabular data. Unlike generalist models such as GPT-4, Yasa can be personalized for proprietary data and applications, thereby providing companies with tailored AI capabilities without requiring in-house expertise. In the future, Reka aims to develop AI that can handle more data types and continuously self-improve. While the company isn't generating revenue yet, it plans to use its funding to acquire computing power from Nvidia and build a business team. Snowflake, one of Reka's early customers and investors, is partnering with the startup to allow its customers to deploy Yasa.
[Blockchain] EU Launch of European Digital Infrastructure Consortium. The European Union has launched the European Digital Infrastructure Consortium (EDIC) to drive blockchain policy and compete with similar initiatives in China. Operational by the end of 2024, the EDIC, led and funded by member states, will work to foster wider blockchain application, beyond its current associations with cryptocurrencies. Initial integration will be with public apps dealing with wallets, digital identities, licenses, and supply chain tracking. Plans also include enabling private applications to build on the EU blockchain. However, questions remain about the technological capabilities of the blockchain, specifically its scalability to handle increasing transaction volumes. Observers note that governments worldwide are recognizing blockchain's role in the future, with speculation that all blockchains may eventually converge into one. Under its new formation, EBSI would allow for international interoperability. Even though there is still not a lot of engagement at that level yet, an EU source close to the project confirmed there are advanced conversations with Canadian authorities about exchanging educational credentials across their blockchain infrastructures.
[Metaverse] Shanghai Reveals Metaverse Masterplan. Shanghai, China's financial capital, aims to develop its culture and tourism metaverse projects into an industry generating an annual revenue of 50 billion yuan (US$6.9 billion) by 2025. Leveraging immersive technologies, the city plans to enhance its cultural and tourism offerings, with the ambition of becoming a global leader in digital innovation. This initiative revolves around creating a metaverse ecosystem that encompasses various cultural and tourism experiences, from AR tourist guides to blockchain-based artworks. By combining this digital immersion with Shanghai's rich cultural heritage and landmarks, the city intends to offer unique and personalized experiences to both domestic and international visitors. The project underscores Shanghai's commitment to maintaining its cultural legacy while embracing digital advances and the potential economic benefits of the metaverse.
[General technology] 'World First' by Beaming 4G LTE Signal from Space to Earth. AST SpaceMobile, a Texas-based company, has successfully transmitted a 4G LTE signal from space that was picked up by regular smartphones, marking a "world first." The company used its BlueWalker 3 satellite for this test and aims to attempt a 5G connection next. The 4G LTE signal was tested in Hawaii on AT&T’s spectrum with Nokia RAN technology, achieving speeds of up to 10.3Mbps. AST SpaceMobile is working towards bringing broadband services to parts of the world with unreliable or non-existent cellular coverage. Their efforts could allow users to text, call, browse the internet, download files, and stream video using a signal beamed from space. The company plans to continue testing and improving this technology. The company it’s not the only one looking to send data down from space — Lynk Global managed to send text messages via satellite in 2020, and Amazon’s satellite-based internet Project Kuiper will send up to 83 satellites into low Earth orbit starting in 2024.
Recommended by LinkedIn
FEATURED: ‘LLM-Assisted Content Analysis: Using Large Language Models to Support Deductive Coding’
By: Robert Chew, John Bollenbacher, Michael Wenger, Jessica Speer, Annice Kim
“Deductive coding is a widely used qualitative research method for determining the prevalence of themes across documents. While useful, deductive coding is often burdensome and time consuming since it requires researchers to read, interpret, and reliably categorize a large body of unstructured text documents. Large language models (LLMs) are a class of quickly evolving AI tools that can perform a range of natural language processing and reasoning tasks. In this study, we explore the use of LLMs to reduce the time it takes for deductive coding while retaining the flexibility of a traditional content analysis. We outline the proposed approach, called LLM-assisted content analysis (LACA), along with an in-depth case study using GPT-3.5 for LACA on a publicly available deductive coding data set. Additionally, we conduct an empirical benchmark using LACA on 4 publicly available data sets to assess the broader question of how well GPT-3.5 performs across a range of deductive coding tasks. Overall, we find that GPT-3.5 can often perform deductive coding at levels of agreement comparable to human coders. Additionally, we demonstrate that LACA can help refine prompts for deductive coding, identify codes for which an LLM is randomly guessing, and help assess when to use LLMs vs. human coders for deductive coding. We conclude with several implications for future practice of deductive coding and related research methods.
“Qualitatively, we found that producing model reasons for coding decisions helps human coders assess model performance, prompt quality, and build trust in predictions. Examples in the Results section demonstrate model generated reasons that helped in identifying hallucinations and reasoning errors and suggests ways in which the codebook can be modified to improve performance. Additionally, we found that model generated reasons can help human coders reflect on their own coding decisions, which can in turn inform revisions to measure definitions. Anecdotally, we find that the process of making the codebook more explicit for the model also tends to help improve the instruction readability and comprehension for human coders. This finding may be less robust if using LLMs that have not gone through the same amount of fine tuning that the GPT-3.5 series of models received, where prompt engineering that more closely mimics next token prediction may be required [46].
Across all data sets, the IRR results suggest that for many cases, GPT-3.5 codes at a level of agreement comparable to human coders. In cases where the model far underperformed human coders, these usually could be detected early with the hypothesis tests of randomness. This finding largely agrees with prior research on LLMs ability to generate annotation for NLP tasks, which suggests that GPT-3.5 or GPT-4 meet or exceed the performance of crowd workers
[ 17 ], and in some instances, also experts [ 18]. Interestingly, our results show more promising LLM results than the prior work using LLM for deductive coding [19], though it’s challenging to draw strong conclusions based on differences in prompting strategies, data sets, metrics, and qualitative coding tasks.”
Read the full paper: here.
Our latest article is out !
Unleashing Potential: How Industry Relations Drive Success for Emerging Tech Businesses ?
In the ever-evolving landscape of emerging technologies, establishing solid industry relations is no longer a luxury—it's a strategic necessity. This article unpacks the concept of industry relations, its importance for tech companies on the cutting edge, and how it can unlock a plethora of opportunities in their respective industries. This influence can help them carve out a competitive advantage, foster collaboration, and gain recognition as industry leaders.
Chief Conversational AI Disruptor @ ChatFusion/ContactLoop | E&Y Entrepreneur of the Yr '08 | $150mn Exit ‘08 | AI Insights for Marketers & Sales Executives
1yAnaïs Ofranc Really good share; keep up the effort.