The essay "The future of the firm" by Hemanth Asirvatham and Elliott Mokski of OpenAI explores three primary scenarios for how AI will integrate into organizational structures, fundamentally reshaping firms, operations, and employee roles. These scenarios suggest AI could be positioned beneath, beside, or between human relationships within a company.

The first scenario, "The Promoting," envisions AI as a "basement" labor force reporting to individual contributors, effectively making everyone a manager. This allows employees to automate their past tasks with AI, focusing on higher-level responsibilities and creative work. This implies a "promotion" of roles rather than layoffs, with entry-level positions being filled by AI, offering younger workers opportunities for higher-scope learning and experimentation.

The second scenario, "The Interleaving," describes AI joining the organizational chart as colleagues, working alongside human teams and managed by existing human managers. In this model, AI acts more like "labor" than a mere "tool," taking on broader functions and adapting to the firm's needs rather than the firm having to learn the technology. The authors highlight an asymmetry where, as AI's scope expands from tasks to teams and even entire firms, its adoption becomes easier.

The third and most complex scenario, "The Layering," proposes an organizational structure where AI is positioned "above" people, creating an "automation sandwich" of human, AI, human. Here, AI functions as a "consultancy" for executives, coordinating and monitoring work to enforce the executive's vision throughout the organization. This could centralize power and reduce information friction, allowing executives a more detailed understanding of their firm, potentially fostering greater trust and freedom for employees while also enabling stricter adherence to top-down directives.

Building on these scenarios, the article addresses four critical questions: "Will firms have more or less red tape?", "Will workers have more or less agency?", "Will firms shrink or widen?", and "Will pay disparity rise or fall?". The authors suggest that cheaper prototyping enabled by AI will reduce red tape, and AI's excellent advice will likely increase worker agency by narrowing the decision-making quality gap. Regarding firm size, AI's ability to streamline communication could lead to both larger and smaller firms, with a prediction of many smaller firms alongside sprawling AI "holding companies."

For pay disparity, the article posits that it could either increase, as those who better manage AI firms beneath them are more highly rewarded, or decrease, as AI provides good advice to everyone, leading to less differentiation. The ultimate outcome, it concludes, will depend on whether humans are primarily providing judgment or following it.

👉 Read the full article on OpenAI

The paper "RAMBO: Leaking Secrets from Air-Gap Computers by Spelling Covert Radio Signals from Computer RAM" by Mordechai Guri details a novel attack method to exfiltrate secret information from air-gapped computer systems. Air-gapped systems are physically isolated from external networks to ensure high security for sensitive data, but this research exposes a critical vulnerability.

In this attack, malware on a compromised computer generates radio signals directly from its memory buses (RAM). These software-generated radio signals can encode sensitive information, including files, images, keylogging data, biometric information, and encryption keys. An attacker can then intercept these transmitted raw radio signals from a distance using Software-Defined Radio (SDR) hardware and a simple, off-the-shelf antenna.

The intercepted signals can subsequently be decoded and translated back into binary information. The paper presents fast modification methods that enable data leakage from air-gapped computers at a rate of 1000 bits per second. Furthermore, the research proposes countermeasures to mitigate this out-of-band air-gap threat.

While this research is not directly related to Generative AI or LLMs, it represents a significant cybersecurity threat in the AI age, especially where highly sensitive data, such as AI models or training datasets, might be stored on air-gapped systems. The discovery of such physical exfiltration channels underscores the increasing importance of robust security measures in an era where data protection is paramount.

👉 Read the full article on arXiv

Unfortunately, the content of this TechCrunch article could not be accessed directly from the provided URL. Attempts to find an alternative source with the specific article content via Google Search were also unsuccessful. Therefore, a detailed summary of the article's content cannot be provided.

However, based on the title, "AI delivers 'zero financial payoff' to most companies, Apollo economist says," it can be inferred that the article likely discusses the realistic challenges of AI adoption, suggesting that investments in AI technology are not yielding expected financial returns for many businesses. This could stem from various factors, such as high implementation costs, integration complexities, or a lack of tangible value creation within existing business processes. This perspective would offer a healthy skepticism towards inflated AI expectations and underscore the need for a cautious approach when companies formulate their AI strategies.

👉 Read the full article on TechCrunch