Apple Researchers Reveal New AI Model 'ReALM' Claiming That It Outperforms GPT-4

It appears that Apple has made significant strides in AI with their new model named ReALM. According to recent reports, ReALM is designed to be smaller and faster than GPT-4, particularly when parsing contextual data. This could make interactions with Siri more efficient, as ReALM is capable of converting context into text for easier processing by large language models.

In a new research paper published on 29th of March, Apple researchers explain how the new Al system, called ReALM (Reference Resolution As Language Modeling), can look at what's on your screen and what you're doing to figure out what you need which means that Siri could understand the context of your questions much better than before, like knowing what's on your screen or what music is playing.

On top of that, Apple researchers claim that the larger models of ReALM outperform GPT-4. If the claims come true, Siri could become much more helpful than ever. The report notes that Apple's ReALM language model purportedly surpasses GPT-4 in "reference resolution," understanding contextual references like onscreen elements, conversational topics, and background entities.

Apple's research suggests that even the smallest ReALM models perform comparably to GPT-4 with fewer parameters, making it well-suited for on-device use. With increased parameters, ReALM substantially outperforms GPT-4. 

Summary of the key findings from Apple's ReALM research paper:

Efficiency: ReALM is designed to be smaller and faster than large language models like GPT-4, making it well-suited for on-device use.

Reference Resolution: The model excels in reference resolution, which is the ability to understand context and ambiguous references within text. This is crucial for interpreting user commands in a more natural way.

Performance: Even the smallest ReALM models performed similarly to GPT-4 with much fewer parameters. When the number of parameters was increased, ReALM substantially outperformed GPT-4.

Image Parsing: Unlike GPT-4, which relies on image parsing to understand on-screen information, ReALM converts images into text, bypassing the need for advanced image recognition parameters. This contributes to its smaller size and efficiency.

Decoding Constraints: ReALM includes the ability to constrain decoding or use simple post-processing to avoid issues like hallucination, enhancing its reliability. 

Practical Applications: The paper illustrates practical applications of ReALM, such as enabling Siri to parse commands like "call the business" by understanding the context, like a phone number displayed on the screen.

Apple's research indicates that ReALM could significantly improve the speed and accuracy of Siri, making interactions with the voice assistant more intuitive and efficient. The company is expected to reveal more about its AI strategy during WWDC 2024.

 
This development is quite exciting as it indicates progress towards more responsive and intuitive AI systems that can better understand and process user commands. It's also a step forward in the integration of AI in everyday devices, potentially enhancing user experience significantly. Apple plans to unveil more about its AI initiatives in June, which could include further applications of ReALM.

PatSeer Unveils Cutting-Edge AI Semantic Search with Custom-Trained GPT Model

Leading Patent Search and Intelligence platform, PatSeer Unveils Cutting-Edge AI Semantic Search with Custom-Trained GPT Model

With this new capability, PatSeer continues to revolutionise the patent research domain by bridging the gap between Boolean and AI-based systems

PatSeer, a leading Indian AI-driven B2B SaaS platform for Intellectual Property (IP) Research and Intelligence is excited to announce a completely new and rebuilt semantic search powered by an advanced GPT-driven language model. The new semantic search replaces its older Natural Language search capabilities in the product by custom-training the underlying LLM model to understand patent semantics. PatSeer's new AI search feature brings a significant improvement in result accuracy on tests conducted across various research sectors, further solidifying their presence as a global company which caters to prominent blue chip firms in India and abroad. PatSeer is also the platform of choice for the Indian Patent Office for their day to day patent examination activities.

With this new development, patent professionals are no longer confined to the contextuality limitations of Boolean searches and can capitalise on PatSeer’s AI capabilities for comprehensive searches. As a platform that’s used by thousands of IP searchers on a daily basis, PatSeer’s complex Boolean search functionality will continue to remain crucial for certain IP projects. However now, with the integration of the new semantic search, PatSeer is offering the best of both worlds, eliminating the need for two separate systems.

Further, with a true Boolean and AI integration, users get unique capabilities that neither can provide independently. For example, one can initiate a semantic search and narrow the results iteratively with Boolean searches and vice versa. It also allows for the semantic re-ranking of Boolean results using a paragraph or a relevant patent. The system also suggests related records to existing results based on the underlying language model’s contextual understanding. For transparency purposes, users can view the closest matching snippets from full text to understand why a record scored higher in their results. The new semantic search is also immediately available as an API for integration into internal systems.

Manish Sinha

“In the many conversations with patent professionals, I’ve been told that Boolean and AI-based patent systems will always be separate tools and I could never accept that,” remarked Manish Sinha, Founder and Chief Technology Officer at PatSeer. “I’m excited to debunk that notion today. PatSeer has become the first to offer a seamless fusion of transformative AI-driven search within a professional patent search database. The days of maintaining access to two separate solutions or paying by the query are over. At PatSeer our commitment to harnessing the power of Deep Learning AI to tackle greater challenges remains stronger than ever. Following the launch of our AI Classifier last year, the advent of our new Semantic Search marks another significant milestone in this journey. We are quite excited with the possibilities here and the road ahead.”

With a commitment to remain at the forefront of AI-driven solutions, PatSeer is paving the way for the future of patent research. PatSeer offers IP experts worldwide patent search with AI technology, faceted filtering, advanced analytics for comprehensive IP business intelligence, an intuitive user experience, and optimised workspaces, and leads the charge in utilising AI within IP research and analysis tools. It aims to establish itself as the pioneer in addressing significant challenges in the field by offering customised AI solutions tailored to meet its clients' unique needs.

Intellectual property (IP) and patent research have become essential elements of today's ever-changing world of technology and innovation. Intellectual Property encompasses various forms, including patents, trademarks, copyrights, and trade secrets. Among these, patents stand out as legal documents that grant their holders exclusive rights to an invention for a specified period, typically 20 years from the filing date. This exclusivity empowers inventors to prevent others from making, using, selling, or importing their inventions without permission.

Patent research involves delving into patent databases to extract useful information about emerging innovations and industry trends. This form of research helps businesses make informed choices regarding product development, market strategies, and intellectual property protection. Beyond its corporate applications, patent research also serves the interests of policymakers, investors and the general public, by providing them with essential data that can influence technological and social advancements.

Edtech Startup Scaler Introduces GPT-4 powered AI Teaching Assistant

Scaler introduces GPT-4 powered AI teaching assistant to enhance learner experience

Since its pilot launch, 6000 learners from its Scaler Academy program have been provided access to the AI Teaching Assistant, enabling round-the-clock query resolution, and the company aims to extend this to all its learners by Q3 of 2023

Scaler (by InterviewBit), one of India's fastest-growing edtech startups, today announced the introduction of its GPT-4 powered AI teaching assistant for learners. By integrating ChatGPT functionalities into the Scaler Academy program, the tech upskilling ‘soonicorn’ aims to reduce the doubt resolution turnaround time of the learners, enhancing their learning outcomes.

This new system will enable comprehensive support for learners round-the-clock, enabling students to learn at their own pace. With the introduction of this solution, Scaler intends to address three potential pain points, namely, understanding problems, identifying optimal problem-solving approaches, and code debugging, thereby ensuring that learners receive immediate assistance at any time of the day.

Once learners identify and input their pain points as a prompt to the teaching assistant, the GPT-powered feature presents them with a suitable response to help them with their queries. Safeguards have been implemented to ensure the teaching assistant doesn't divulge the actual solution to the learner. Instead, prompts are presented that help them understand the problem better and figure out the answer independently. This approach promotes a hands-on and self-driven learning experience, where learners can develop their problem-solving skills and gain a deeper understanding of concepts. Through the implementation of 'prompt engineering,' Scaler aims to ensure that learners receive the best responses from ChatGPT, empowering them in their journey to overcome challenges and resolve doubts effectively.

Abhimanyu Saxena & Anshuman Singh - Co-Founders, IB & Scaler

In addition to this, Scaler has also introduced a 'Text Help Request' (THR) feature that enhances the learner experience by providing immediate and high-quality doubt resolution. Leveraging advanced machine learning algorithms such as BERT and Distilled BERT, the THR feature ensures accurate and contextually relevant guidance. Scaler's system delivers the most appropriate THR recommendations based on learners' specific queries by utilising BERT for text similarity matching and for computing text embeddings. Parameters such as Help Request and Teaching Assistant ratings, recency of resolved HRs, and a weighted scoring system contribute to the accuracy and quality of the recommendations. Within a week of its implementation, Scaler is already seeing a 30% better resolution of help requests raised by learners on average, an all-time low.

Abhimanyu Saxena, Co-founder of Scaler & InterviewBit, said, "While being mindful and empathetic towards the busy schedules of working professionals who undertake tech upskilling courses, it is crucial to provide them with an outcome-driven and high-quality learning experience at every stage of their journey. Previously, our Scaler learners relied solely on teaching assistants who were available for 15 hours a day to address their doubts. However, with the launch of our GPT-4 powered AI teaching assistant and the THR feature, learners can now have their doubts and queries addressed instantly. These new features greatly boost productivity for our learners, aiding them with all the means necessary to work their way around complex concepts and problem-solving challenges. By leveraging the power of AI and ML, Scaler is reshaping the landscape of tech education, empowering learners to achieve their goals efficiently and effectively."

Since the pilot launch a month ago, 6000 learners from its Scaler Academy program have been provided access to the AI Teaching Assistant, with the feature already significantly impacting learners' experiences. With over 33% more learners utilising the tool to solve problems independently without the intervention of a teaching assistant, Scaler has witnessed a significant uptick in learner satisfaction. The learner analytics also observed that learners who accessed the ChatGPT-powered AI teaching assistant were 10% to 20% less likely to raise help externally and could solve problems by themselves - the key idea being building a problem-solving attitude among learners towards software engineering that eventually leads to better outcomes within the program as well as on the job. The majority of exposed learners have rated the support as 5* and are glad they are able to get support from AI to code better!

Over the next three months, Scaler aims to make both the AI Assistant and the Text Help Request features accessible to all their 35,000 learners in the Scaler ecosystem.

About Scaler:

Launched in 2019, Scaler is India's leading tech education company that upskills working professionals and educates aspiring engineering students. Scaler's industry-vetted curriculum provides solutions to real-world challenges addressing the changing dynamics of the technology industry through two flagship programmes: Scaler Academy and Scaler DSML.

Scaler believes in creating real-life impact by focusing on 'impact-driven' tech talent. Enrolled learners are mentored and taught by leaders and subject-matter experts working with leading organisations, including MAANG companies. Scaler has generated enviable career outcomes in a short period, and on average, its learners see a 4.5x RoI (return on investment) and salary hike of approximately 126%.

To further redefine tech education, Scaler has launched Scaler Neovarsity - an online university that offers an outcome-focused Master's Programme accredited with ECTS and Scaler School of Technology - a 4-year residential UG program in Computer Science. Scaler is also the only edtech player in the country to establish 'Scalerverse', a distributed campus to add to its Neovarsity vision for learners, teachers, and mentors, encouraging holistic learning and development. Scaler Enterprise is the B2B arm that focuses on building relationships with domestic and global organisations to provide them with industry-ready talent. Over 800 companies have worked with Scaler for their tech recruitment needs.

The startup's parent firm, InterviewBit, is featured on the Financial Times's Asia Pacific High Growth Companies 2021, 2022 and 2023 ranking. Scaler is backed by marquee global investors like Peak XV Partners (formerly Sequoia Capital India), Tiger Global, and Lightrock India and has expanded its footprints across India and the US.