HyperWrite’s Reflection 70B, a new large language model (LLM) based on Meta’s Llama 3.1-70B Instruct, has been introduced by Matt Shumer, CEO of HyperWrite. Shumer claims that Reflection 70B is now the most powerful open-source AI model in the world, outperforming Meta’s Llama series and competing with top commercial models. What sets Reflection 70B apart is its innovative self-correction capability, allowing the model to recognize and fix its own mistakes during inference, enhancing its accuracy in various tasks.
Reflection 70B has been tested on multiple benchmarks, including MMLU and HumanEval, using LMSys’s LLM Decontaminator to ensure clean, reliable results. These benchmarks demonstrate the model’s ability to surpass its competitors in terms of performance. Users can interact with the model through a demo on a dedicated website, but the overwhelming demand has caused traffic issues, with Shumer’s team working to scale up GPU resources to meet the surge.
A key feature of Reflection 70B is “reflection tuning,” which allows the model to detect errors in its reasoning and correct them in real time. The system uses special tokens to mark reasoning steps, enabling more structured interaction and accurate outputs. This makes it particularly useful for tasks requiring precision, such as comparing numbers or counting characters in words, areas where many other models often fail.
Shumer also revealed plans for an even more powerful model, Reflection 405B, set to release soon, which is expected to outpace top closed-source models like OpenAI’s GPT-4. In addition, HyperWrite intends to integrate Reflection 70B into its main AI writing assistant, with details on this integration expected soon.
The rapid development of Reflection 70B was made possible by Glaive, a startup specializing in generating synthetic data for AI model training. Glaive’s technology allowed the HyperWrite team to quickly create high-quality datasets tailored to their needs, reducing the model’s training time to just three weeks. Shumer credited Glaive for this efficiency, highlighting how synthetic data can accelerate AI development.
HyperWrite, originally founded as Otherside AI in 2020, has grown significantly since its inception. Initially launched as a tool for crafting emails, HyperWrite has expanded to provide a range of AI-driven services, with two million users as of late 2023. The company’s cautious approach to accuracy and safety in AI development mirrors the reflection capabilities embedded in its latest models.
Looking forward, Shumer’s ambition is for Reflection models to set new standards for open-source AI, challenging established players in the industry like OpenAI, Anthropic, and Microsoft. With Reflection 405B on the horizon, the balance of power in generative AI might be shifting towards open-source innovations.
Concerns about an impending fourth AI winter are rising as doubts emerge over whether artificial intelligence will deliver enough practical benefits to justify its high costs. Recent reports from Goldman Sachs and other research bodies highlight this skepticism. However, a solution has been present for some time—engineered intelligence, a concept emphasizing the practical application of AI through engineering principles.
Traditionally, scientific breakthroughs in fields like chemistry and physics are first made in laboratories and then transferred to engineers to develop real-world applications. This process ensures that discoveries are turned into practical solutions. However, AI lacks a similar transition mechanism. Instead of a dedicated discipline for applied AI, organizations often hire data scientists—primarily researchers—to work on developing practical AI solutions. This mismatch has contributed to a significant failure rate, with 87% of AI projects not reaching successful outcomes.
Engineered intelligence, or intelligence engineering, is emerging as a new field focusing on applying AI research in practical settings, much like how chemical or mechanical engineers utilize scientific discoveries. This discipline allows experts, scientists, and engineers to develop intelligent solutions without needing to become data scientists. By reestablishing research-to-engineering pipelines and forming partnerships with academic institutions and technology vendors, industrial organizations are setting the stage for engineered intelligence. This approach mirrors how breakthroughs in other scientific areas are handed off to specialized engineers.
With intelligence engineering, AI research can be transformed into breakthrough applications, yielding tangible value and producing outcomes that might not have been identified by data scientists alone. The process facilitates creating value-driven AI solutions that are feasible and safe for production use, contributing to meaningful advancements across various industries.
To introduce intelligence engineering into an organization, practical experience is crucial. Here’s a five-step guide that differs from traditional AI implementation methods:
By adopting intelligence engineering, organizations can expand their capabilities, moving beyond existing expertise to identify new opportunities. This approach enables safe and practical value creation, both within organizations and across broader ecosystems. As more industries and educational institutions develop programs focused on engineered intelligence, the resulting innovations will unlock unrealized economic and societal benefits, paving the way for new job categories and a surge in value creation.
Brian Evergreen, author of Autonomous Transformation: Creating a More Human Future in the Era of Artificial Intelligence, and Kence Anderson, author of Designing Autonomous AI, both emphasize the potential of engineered intelligence to redefine AI’s impact in practical, everyday applications.
The rise of “emotion AI,” which aims to equip artificial intelligence with the ability to understand human emotions, is becoming a notable trend in business software, according to PitchBook’s recent Enterprise SaaS Emerging Tech Research report. This technology is seen as a step beyond sentiment analysis, promising more nuanced interpretations of human interactions by using multimodal inputs like visual, audio, and psychological data. Despite its potential, the effectiveness and ethical implications of emotion AI remain questionable.
The concept behind emotion AI is straightforward: as businesses increasingly rely on AI for customer service, sales, and other interactions, these AI bots need to distinguish between different emotional cues, such as anger and confusion. Emotion AI intends to make AI assistants more human-like in their responses by analyzing various signals, from facial expressions to voice tones. Major cloud providers, including Microsoft and Amazon, already offer services with emotion AI capabilities, making these tools more accessible to developers.
Derek Hernandez, a senior analyst at PitchBook, notes the growing importance of emotion AI with the proliferation of AI assistants and automated human-machine interactions. Hernandez highlights the role of cameras, microphones, and wearable devices in capturing the necessary data for emotion detection. This growing interest has spurred investment in startups like Uniphore, MorphCast, Voicesense, and others, which focus on developing emotion AI technologies.
However, the push toward emotion AI brings with it significant challenges. Critics argue that the technology might be inherently flawed. Research published in 2019 suggests that human emotions cannot be accurately determined by facial movements alone, challenging the basic premise of emotion AI. Moreover, regulatory concerns, such as those outlined in the European Union’s AI Act, which restricts emotion detection in specific contexts, could limit its application. U.S. state laws like Illinois’ Biometric Information Privacy Act (BIPA) further complicate the use of biometric data without explicit consent.
The debate around emotion AI offers a glimpse into the potential future of AI in the workplace. While emotion AI could enhance customer service, sales, and HR tasks by making interactions more personalized and empathetic, it raises questions about privacy, ethical implications, and the actual effectiveness of such technology. As companies continue to embed AI across various aspects of business operations, the success and acceptability of emotion AI will likely depend on addressing these challenges.