Emerging Tech & AI - 7th Edition

Welcome to the 7th Edition of?the Emerging Tech & AI Newsletter!

This newsletter's goal is to help you stay up-to-date on the latest trends in emerging technologies. Subscribe to the newsletter today and never miss a beat!

Subscribe to the newsletter here.

Here's what you can expect in each issue of the Emerging Tech & AI Newsletter:

• A summary of the top AI / emerging technology news from the past week
• Introductory details of a key topic in AI or any other emerging technology (We review different metrics that can be used to Measure LLM Performance?this week)
• A primer on an emerging technology (We review the many use cases of NFTs)
• Examples of how AI tools are being used ( We explore How AI can help in improving CX? )

Let's explore the future of technology together!

Last Week in AI/Emerging Tech

The field of AI is experiencing rapid and continuous progress in various areas. Some of the notable advancements and trends from the last week include:

Big Tech in AI:

1. Microsoft and Mercy collaborate to empower clinicians to transform patient care with generative AI.
2. Microsoft is rolling out AI-powered Paint Cocreator to Insiders.
3. Microsoft launched AutoGen to enable next-generation large language model applications.
4. Google introduced Google-Extended to let anyone block Bard, and Vertex AI via robots.txt.
5. Bard AI gets smarter: Interacts with Google services to deliver personalised experiences.
6. Meta announced AI chatbots with 'personality'.
7. Meta is putting AI chatbots everywhere.
8. Meta CEO Mark Zuckerberg unveiled the redesigned Ray-Ban Meta smart glasses.
9. Meta AI released Emu, a quality-tuned latent diffusion model for generating highly aesthetic images
10. AWS announced the general availability of Amazon Bedrock and powerful new offerings to accelerate generative AI innovation.

Funding & VC Landscape:

1. Amazon will invest up to $4 billion in Anthropic and have a minority ownership position in the company.
2. Chinese AI chip maker Enflame raised $274 million in Series D.
3. AI-driven market intelligence platform AlphaSense locked up a $150 million Series E led by Bond Capital at a $2.5 billion valuation.
4. London-based AI platform Flawless raised $2.2M to free up operations manager’s time.
5. Kolena, a startup building tools to test AI models, raised $15M.
6. AI chip company Kneron raised $49M to scale up.
7. Nexusflow raised $10.6M to build a conversational interface for security tools.
8. Correcto grabbed $7M to build out its ‘Grammarly for Spanish’.

Other AI news:

1. Zapier launched Canvas, an AI-powered flowchart tool.
2. SAP launched Joule, the generative AI assistant, to interact with SAP business systems.
3. Paygeon Launched AI-Powered Financial Platform for SMBs.
4. US Intelligence Agencies Are Putting the AI in CIA.
5. Abacus AI released a larger 70B version of Giraffe.
6. Cloudflare launched Workers AI (an AI inference as a service platform), Vectorize (a vector Database) and AI Gateway with tools to cache, rate limit and observe AI deployments.

Liked the news summary. Subscribe to the newsletter to keep getting updates every week. Check the comments section on the LinkedIn article for links to the above news.

Key Metrics for Measuring LLM Performance

There are several ways to measure the performance of large language models (LLMs). Some of the most common metrics include:

1. Quantitative Metrics:Perplexity: Perplexity is a measure of how well an LLM predicts the next word in a sequence. It is calculated by averaging the negative log-likelihood of the next word across a set of test sequences. Lower perplexity indicates better performance.Accuracy: Measure the model's accuracy on specific tasks such as text classification, sentiment analysis, or question-answering. High accuracy suggests better performance.F1 Score: For tasks like text classification or named entity recognition, the F1 score is a useful metric that balances precision and recall.BLEU score: The BLEU score is a measure of the similarity between a generated text and a reference text. It is calculated by comparing the n-gram overlap between the two texts. A higher BLEU score indicates better performance.ROUGE: ROUGE is a set of metrics used for evaluating the quality of summaries. It compares the generated summary with one or more reference summariesCosine Similarity: Cosine similarity helps us understand how similar two vectors are. We can compare the output of LLM models against a reference prompt and answer and calculate cosine similarity against the reference answer to measure performance.
2. Downstream Task Performance:Evaluate the model on specific downstream tasks that it's designed for, like translation, summarization, text generation, or chatbot functionality. You can use task-specific metrics to measure performance in these contexts. The Mosaic Eval Gauntlet is MosaicML’s technique for evaluating the quality of pre-trained foundation models. The Eval Gauntlet encompasses 34 different benchmarks collected from a variety of sources and organized into 6 broad categories of competency that we expect good foundation models to have.
3. Human Evaluation:Human evaluation is a qualitative assessment of the quality of an LLM's output. Human evaluators typically rate the generated text on factors such as fluency, coherence, factual accuracy, and overall quality. One of the techniques that can be used for evaluation at the start of development is eyeballing.
4. User Satisfaction:Measure user satisfaction and engagement when interacting with applications powered by LLMs. Surveys, feedback forms, or user studies can help gather this information.
5. Bias and Fairness Analysis:Evaluate the model's bias and fairness by examining its responses to different demographic groups. Tools like Fairness Metrics can be used to quantify bias.
6. Few-Shot and Zero-Shot Performance:Assess how well the LLM performs in situations where it has limited or no training data, as this is one of the strengths of many LLMs. Tools like LM-Eval can be helpful here
7. Transfer Learning:Measure the efficiency of transfer learning from pre-trained LLMs to specific tasks. You can evaluate how much fine-tuning or task-specific training is needed for good performance. Tools like MosaicML can help evaluate LLMs on in-context learning tasks (LAMBADA, HellaSwag, PIQA, and more).
8. Resource Efficiency:Evaluate the computational resources required to run the model, including inference time and memory usage.
9. Ethical Considerations:Assess the ethical implications of the LLM's performance, including the potential for misuse or harmful outputs.
10. Benchmarking:Compare the LLM's performance against other models, both past and present, to understand its relative strengths and weaknesses. Some of the benchmarks are Glue, SQuAD2.0, SNLI,BIG Bench etc

It's important to note that LLM performance evaluation can be complex, and a single metric may not provide a comprehensive picture. Often, a combination of these methods and metrics is necessary to accurately assess an LLM's overall performance. Additionally, the choice of metrics and evaluation criteria should align with the specific use case or application for which the LLM is being employed.

Want us to explore more? Let us know in the comments and we will plan a full blog on Technologia.

The Many Use Cases of NFT

If you want to know more about NFT, check out this link. This section is a partial reproduction of the original article on Technologia.

Over the last few years, NFTs have been used in many areas, some of which are mentioned below:

1. Digital Art /Collectibles — Any digital art can be minted as an NFT and it then becomes a unique digital art as the hash associated with each NFT is unique. Many traditional art auction houses like Christie’s have entered the NFT world in the last few years. NFTs are designed to give ownership of work only and do not restrict making any copies. Just like we can find print copies of The Last Supper but there is only one original, there can be only one original NFT based digital art. Many individual artists have also sold their digital art as NFTs. NFTs have been used to create Digital Collectibles by many well-known brands and franchises. Many digital collectibles based new communities have also started in the last 2 to 3 years. Digital collectibles have also been used in games.
2. Digital version of real-world items — Many real-world items can be replicated in the digital world using NFTs. For instance, tickets to events, legal documents/deeds and so on
3. Gaming — In addition to providing in-game items /collectibles, NFTs can be used to create new revenue models in-game economy. Imagine trading a farm created in-game like Farmville or selling a particular level that you have reached in a game to someone else who wants to play that. New avenues of generating money by doing something you already like. Collectibles can be resold to new players generating revenue for both gamers and game creators.
4. NFT Domains— As of now, IP addresses are mapped with domain names to find any website. In the web3 world, we have equivalent services like Ethereum name service which provide alternate website names powered by NFTs. NFT domains are new web extensions that are linked to the blockchain via smart contracts. NFT domains map hexadecimal wallet addresses into human-readable form and enables censorship-resistant websites.
5. Authentication — Since NFTs are unique and cannot be reproduced, they can be used for verification and authentication also. Everledger uses NFTs to solve this use case in the supply chain. Identity credentials like driver licenses, course certifications etc can also be issued as NFT in future.
6. Proof of ownership — Cryptographic proof of ownership can be used as a key in future to unlock cars, homes or any other physical assets. NFTs can be used to tokenise ownership of physical assets giving way to fractional ownership of real estate etc. As the ownership is settled by NFTs, they can also be used in decentralized loan applications also.
7. Royalties — Creators can earn some commission on each future sale by using NFTs. Creators can have a smart contract that triggers an auto payment to the original creator whenever the NFT change hand. So a music video or digital audio will earn recurring revenue if it’s transferred. NFTs will also allow creators to directly sell their content to a mass audience without any intermediatory.

There are no limits to what the creative human mind can up with! The use of NFT is not limited to the above uses cases only but these are the common ones.

How AI can help in improving CX

Artificial Intelligence (AI) can play a significant role in improving Customer Experience (CX) across various industries. Here are several ways AI can enhance CX:

1. Personalization:AI can analyze customer data and behavior to create highly personalized experiences. It can recommend products, services, and content tailored to individual preferences, increasing engagement and satisfaction.
2. Chatbots and Virtual Assistants for Improving customer service responsivenes:AI-powered chatbots and virtual assistants can provide 24/7 support, answer common customer queries, and assist with tasks like booking appointments or making reservations, improving response times and accessibility.
3. Sentiment Analysis:AI can analyze social media posts, customer reviews, and feedback to gauge customer sentiment. This information can help companies identify issues, trends, and opportunities for improvement.
4. Predictive Analytics:AI algorithms can forecast customer behavior and preferences. This can help businesses anticipate needs and optimize inventory management. This information can also be used to proactively reach out to customers and offer assistance, or to develop new products and services that meet the needs of the customer base.
5. Recommendations:AI-driven recommendation engines can suggest products, services, or content based on a customer's past interactions, increasing cross-selling and upselling opportunities.
6. Automation:AI can automate routine tasks and processes, reducing the need for manual intervention. This can lead to faster response times and increased efficiency in resolving customer issues.
7. Customer Insights:AI can analyze large datasets to provide actionable insights about customer behavior, allowing businesses to make data-driven decisions to enhance CX.
8. Voice Recognition:AI-powered voice recognition technology can improve CX in call centers and with voice-controlled devices, making interactions smoother and more accurate.
9. Fraud Prevention:AI can detect and prevent fraudulent activities, such as unauthorized transactions, protecting both businesses and customers from potential harm.
10. A/B Testing and Optimization:AI can help businesses test different strategies, content, and layouts to identify what resonates best with their audience, leading to continuous improvement in CX.
11. Language Translation:AI-driven translation tools can break down language barriers, allowing businesses to communicate with customers in their preferred language.
12. Accessibility:AI can be used to improve digital accessibility, making websites, applications, and content more user-friendly for people with disabilities.
13. Self-Service Options:AI-powered self-service options like knowledge bases and FAQs can empower customers to find answers to their questions independently, reducing the need for direct support.

By leveraging AI in these and other ways, businesses can enhance CX, increase customer satisfaction, and build long-lasting relationships with their customers. However, it's crucial to implement AI in a thoughtful and ethical manner, respecting privacy and ensuring transparency in AI-driven interactions.

Interested in knowing more. Let us know in the comments and we will plan a full blog on Technologia.

Disclosure: Some content in the article was written with the help of Google Bard and ChatGPT.

Thanks for reading. See you next week!