The article discusses the potential for an "arms race" in the use of AI tools in litigation. It explores how AI can provide advantages to law firms in terms of improving their odds at trial and inferring complex information. The article also mentions the possibility of different economic models in billing and the impact of AI on the billable hours of litigators. The Legal Innovators UK Conference is mentioned, which will have a dedicated day for litigation in 2025.

Amazon is facing a lawsuit alleging mistreatment of workers in its efforts to catch up in the AI race. The suit claims that the company disregarded worker rights in its pursuit of technological advancement. The case highlights the challenges and ethical concerns surrounding the rapid development of AI technology and its impact on workers.

The AI hype bubble is deflating as evidence suggests that the industry is facing challenges in terms of regulatory issues, legal challenges, and profitability. Despite continued investment and growth in AI companies, widespread adoption and business success are still a long way off. Start-ups are trying to find ways to monetize AI technology, but there are concerns about the reliability and limitations of generative AI. Additionally, the industry is grappling with the need for more electricity to power the data centers required for AI training.

Microsoft, OpenAI, and Google are developing conversational artificial intelligence technology with new features to handle complex tasks. Microsoft is creating software to automate actions like creating invoices and rewriting code, powered by OpenAI's technology. These capabilities will be announced at Microsoft's annual Build developer conference next month.

GCs are relying on legal ops as a solution to their increasing workload. This article discusses how legal operations can help alleviate the pressure on general counsels and improve efficiency in legal departments. The article provides insights into the role of legal ops and its impact on the overall legal function. For more details, refer to the article: "Stretched to the Limit, GCs Turn to Legal Ops as Escape Valve" on Corporate Counsel.

The 2024 AI Index Report provides comprehensive insights into the current state of artificial intelligence (AI). Key takeaways include AI surpassing human performance in certain tasks, industry dominating frontier AI research, the increasing cost of training state-of-the-art AI models, the United States leading in top AI models, the lack of standardized evaluations for responsible AI, the surge in investment in generative AI, AI's impact on productivity and quality of work, AI's contribution to scientific progress, the rise of AI-related regulations, and growing awareness and nervousness about AI's impact.

The article discusses the uncertainty surrounding the development and location of artificial general intelligence. It emphasizes that even experts are unsure about its progress and highlights the speculative nature of predictions. The article provides a link to the full content for further reading.

Companies are interested in harnessing the potential of generative AI but often lack guidance on how to do so. This playbook offers a framework for companies to prioritize use cases and begin development. It covers developing solid use cases, utilizing proprietary data, selecting the appropriate generative AI model, mitigating risks, optimizing user experience, and concludes with a focus on solutions benefiting external customers.

Companies facing increasing class actions due to rising employee activism, according to Corporate Counsel. URL [https://www.law.com/corpcounsel/2024/04/09/companies-deluged-by-class-actions-as-employee-activism-surges/]

AI language models reflect biases and building an unbiased model is technically impossible. Elon Musk is working on "TruthGPT" as an alternative. OpenAI and Google's chatbots have been accused of political bias. Bias in AI language models is a complex problem with no easy fix. Companies should be more transparent about biases. Worldcoin's "World ID" is being investigated for collecting biometric data. Porcha Woodruff is the first woman wrongfully arrested due to facial recognition. AI chatbots spreading lies raise concerns. AI startup hype may be losing consumer interest. Meta plans to launch chatbots with different personalities.

Tech giants harvest data for AI, compromising privacy. Article explores their methods and implications. [URL: https://www.nytimes.com/2024/04/06/technology/tech-giants-harvest-data-artificial-intelligence.html [https://www.nytimes.com/2024/04/06/technology/tech-giants-harvest-data-artificial-intelligence.html]]

Investors in talks to help Elon Musk's xAI raise $3 billion for funding, according to a report by The Economic Times.

This article discusses prompt engineering for AI models like ChatGPT. It explains the importance of prompt engineering and how it can improve the output of AI models. The author also mentions that prompt engineering will continue to be relevant even as AI models advance. The article provides five principles of prompting and emphasizes the need for direction in working with AI and human coworkers. Overall, it highlights the significance of prompt engineering in maximizing the potential of AI models.

Disruptive technologies like AI are creating a challenge for organizations, as they struggle to balance security and employee productivity. Employees are embracing generative AI, remote work, and unapproved applications, putting pressure on IT and security teams. Finding the right balance is difficult, with security professionals feeling pulled in conflicting directions. Shadow IT is also a concern, with employees using unapproved apps and tools. The rise of generative AI adds to the tension, with security concerns around data privacy and phishing attempts. 22% of employees admit to violating company rules on the use of generative AI.

Saudi Arabia is pushing for AI power with direct investments in startups and the creation of a potential $40 billion AI fund. The kingdom aims to develop homegrown AI technology by encouraging startups to establish a presence in Saudi Arabia. This move signifies a shift from relying on venture capitalists to invest on its behalf. The megafund, if launched, would rival Saudi Arabia's stake in SoftBank's Vision Fund.

China plans to disrupt elections in the US, South Korea, and India using AI-generated content, according to Microsoft. The tech firm expects Chinese state-backed cyber groups, along with North Korea, to target high-profile elections in 2024. China will create and distribute AI-generated content through social media to benefit their positions in these elections. While the impact of such content is currently low, Microsoft warns that it may become more effective in the future. China previously attempted an AI-generated disinformation campaign during the Taiwan presidential election.

Tech companies are in a race to buy AI training data, including photos and videos, to train generative AI models. Photobucket, a once-popular image-hosting site, is in talks with multiple tech companies to license its 13 billion photos and videos. The negotiations reveal a growing data market in the quest to dominate generative AI technology. Google, Meta, and Microsoft-backed OpenAI initially used free scraped data from the internet, but they are now quietly paying for content behind paywalls and login screens. The AI data market is estimated to be around $2.5 billion and could reach $30 billion in a decade.

AI chatbots often produce inaccurate outputs, a phenomenon known as "hallucination." This stems from their fundamental construction as autocomplete tools and the limitations of their training data. Fact-checking systems and verifiers are necessary to minimize hallucination. Researchers are working on detecting and correcting errors in AI-generated text. Specialized language models paired with reliable information sources can improve accuracy. In the future, AI tools designed for specific contexts may replace all-purpose models, while generalist chatbots serve as creative partners.

Elon Musk is increasing engineer pay in the AI talent war, according to the Wall Street Journal. The article discusses the impact of Musk's efforts and provides a link for more information.

Despite concerns, legal professionals continue to use generative AI for research, with legal research being the most common use case. The viability and usability concerns surrounding AI-powered legal research are being explored by law firms, legal departments, and courts. The adoption of technology in the legal industry, such as eDiscovery, has faced similar scrutiny in the past. Results from the 2024 Legal Operations and Technology survey will be available on Bloomberg Law's Surveys, Reports, and Data Analysis page.

Google Books is indexing low quality, AI-generated books that could impact Google Ngram viewer. These books, found using a specific search query, include both AI-generated and human-written content. The AI-generated books often contain outdated information and lack depth. Google stated that these books do not currently affect Ngram viewer results, but their inclusion in the future could change the meaning of the data. The presence of AI-generated content online, including books, raises concerns about the reliability of human-made culture. Google has not indicated whether they will filter out AI-generated books or implement a policy regarding them.

AI has lost its magic as generative AI tools have become mundane and less imaginative, transitioning into practical utilities. The technology's rapid adoption and sustained user growth have contributed to its transition into something unexciting. The author reminisces about the early days of using AI for creative purposes but acknowledges that the fading magic and nostalgia are indicative of the technology's power and influence.

Generative AI is seen as a waste of time and money due to its limited social benefits and significant social costs. It relies on computing capacity and energy, leading to environmental concerns and rising energy prices. The capital investment in computing infrastructure and the funding diverted to AI also raise concerns. Additionally, the proliferation of AI-produced "data" can lead to unforeseen consequences and the need to adapt institutions to address the impact of generative AI. Social scientists should be involved in making collective decisions about the future of generative AI.

A fake law firm called Commonwealth Legal is sending fake copyright infringement notices as an SEO scam. The scam involves demanding website owners to place a link to a specific website or face legal action. The law firm is not real, and the images of its "lawyers" are AI-generated. This scam is a backlink SEO scam aimed at improving Google rankings. The scam targets well-established websites and uses the guise of a legal threat. This tactic indicates the rise of insidious SEO scams.

OpenAI introduces new customization options for GPT-4 API, allowing enterprise customers to improve AI accuracy. The options include connecting to third-party platforms, saving fine-tuned models, and a user interface for performance comparison. OpenAI COO Brad Lightcap states that 2024 is the "year of the enterprise" with significant growth in real business applications. Assisted fine-tuning and the Custom Models program are also announced. OpenAI has collaborated with startups and established organizations to bring AI-powered services to the mainstream.

Big Tech companies create consortium to address concerns of AI taking over jobs. Read more on TechCrunch.

S&P Global has launched S&P AI Benchmarks by Kensho, a benchmarking tool for evaluating the performance of large language models in financial applications. The tool ranks models based on their ability to handle tasks such as quantitative reasoning and data extraction from financial documents. S&P Global aims for the tool to become an industry standard and fuel innovation in the FinAI space. The lack of standardized benchmarks has made it challenging for organizations to assess the suitability of different models for their specific use cases.

Excessive workloads are causing high turnover rates in corporate legal departments, leading to a snowball effect. This article discusses the impact of heavy workloads on in-house lawyers and the need to address this issue to improve retention. Source: Corporate Counsel.

Business schools are increasingly focusing on AI, as it becomes a crucial aspect of the business world. [URL: https://www.wsj.com/tech/ai/generative-ai-mba-business-school-13199631?mod=mhp#comments_sector [https://www.wsj.com/tech/ai/generative-ai-mba-business-school-13199631?mod=mhp#comments_sector]]

Large language models (LLMs) are evolving into versatile, semi-autonomous agents. Prompt engineering activates their conversational and task-driven abilities. With additional components like knowledge, memory, and interfaces, LLM agents can engage in personalized dialogues and execute workflows. The quality of prompts is crucial for directing these systems effectively and unlocking their full potential.

The environmental impact of AI production is being considered, with ChatGPT consuming 500ml of water for every 50 texts generated. Microsoft's water consumption has spiked due to AI work. The global AI demand may withdraw billions of cubic meters of water, equivalent to half of the UK's total annual water usage. Reducing computational intensity through quantization can help reduce environmental impact. The cost of accessing AI models like ChatGPT may increase as businesses pay for usage. The environmental impact of AI is a growing concern, but cost pressures may drive efforts to mitigate it.

Microsoft splits Teams and Office apps globally, following EU split. Unbundling due to pressure. Active status.

A visual introduction to Transformers and what a GPT is. Includes a YouTube video explaining the topic. Created on April 6, 2024, and still active.

Washington state judge blocks use of AI-enhanced video as evidence in a possible first-of-its-kind ruling. The judge described the technology as novel and relying on "opaque methods." The ruling comes as AI and its uses, including deepfakes, are evolving, and lawmakers are grappling with the potential dangers. The case involved a man accused of a triple murder who sought to introduce cellphone video enhanced by machine learning software. The enhanced video was deemed inaccurate, misleading, and unreliable by the prosecutor's office.

The article discusses the false promise of AI and its limitations. It emphasizes the need for realistic expectations and highlights the potential risks and challenges associated with AI technology.

Generative AI is used in various areas of life, including technical assistance, content creation, support, education, creativity, and decision making. (66 tokens)

Conversations about the future of AI often focus on apocalyptic scenarios, but the real threats lie in the immediate impacts of AI on jobs and workplaces. AI is already supercharging productivity and changing how we work and learn. Educated and highly paid workers are most exposed to AI, and organizations must make important decisions about how to use AI to benefit their workforce. Instead of fearing AI, we should focus on reorganizing work around AI to make it more meaningful and high value. The choices we make about AI usage will shape its impact on our lives.

US Congress bans use of Microsoft AI Copilot due to concerns about generative AI chatbots. The ban follows the formation of a bipartisan task force on AI and aims to address potential risks in federal agency adoption of AI. Microsoft plans to deliver AI tools that meet federal government security requirements later this year. The ban comes after concerns about fake content, misinformation, and job elimination. Tech companies have also vowed to take measures against AI misuse in elections. Generative AI tools have surged in popularity in recent years.

AI companies are facing a shortage of training data as the internet is not big enough to provide all the data they need. Some companies are exploring alternative sources such as publicly-available video transcripts and AI-generated "synthetic data." However, training AI models on AI-generated data has raised concerns about model collapse. Despite the data shortage, researchers believe breakthroughs can address the issue. Another solution suggested is for AI companies to stop focusing on creating bigger models due to the environmental impact and resource consumption.

ChatGPT, AI language models, and their impact on law. Article ID: 660b2a174b9dd1c101bd0eb2. Created: April 2, 2024. Last edited: April 11, 2024. Status: Active. Article Link [https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4779694]

The AI industry is facing a legal iceberg, with potential implications and challenges ahead. For more details, refer to the article at the provided URL.

New York City's AI-powered chatbot, designed to help businesses navigate government regulations, has been providing incomplete and inaccurate information, sometimes even encouraging businesses to break the law. The chatbot, powered by Microsoft's Azure AI services, has been misinforming users about housing policies, worker rights, and other regulations. The city acknowledges the issues and plans to improve the chatbot. The incident highlights the challenges and limitations of using AI chatbots in providing accurate and reliable information.

Up to 8 million UK jobs at risk from AI unless government acts, finds IPPR.

AI could remove almost 8 million UK jobs, with women, youngsters, and low-paid workers most at risk, warns IPPR report. Government action needed to prevent a "jobs apocalypse" and harness AI's potential for economic growth.

This article explores the use of large language models for relevance judgments in legal case retrieval. It proposes a few-shot workflow that integrates expert reasoning to enhance accuracy. The study shows that reliable relevance judgments can be obtained using this workflow and demonstrates the potential to improve existing retrieval models.

AI is advancing rapidly, outpacing regulatory efforts. Companies are adapting to cope with this challenge.

Anthropic's Claude 3 Opus surpasses OpenAI's GPT-4 on Chatbot Arena for the first time, marking a notable moment in the history of AI language models. The victory signifies a shift in the leaderboard, with Anthropic's models taking the lead. OpenAI's GPT-4, although updated multiple times, is over a year old. The competition in the AI language model space is expected to continue, with OpenAI preparing to release a major successor to GPT-4 Turbo.

Silicon Valley tech leaders are organizing a conference in Washington to drum up more hawkishness towards China's progress in artificial intelligence (A.I.). The event aims to lobby against onerous A.I. regulations, request more government spending on A.I. technology and research, and relax immigration restrictions for A.I. experts. The conference highlights an area of agreement between Washington and Silicon Valley amid the U.S.-China tech and economic cold war. Over 100 tech chiefs, investors, and lawmakers are expected to attend the event.

Apple plans to unveil its AI strategy at the June 10 Developers Conference, focusing on new proactive features for users. Talks with Google and OpenAI for generative AI services. iOS 18 update will be a major overhaul of iPhone's software.

New research links memory loss and declining grades in students who heavily rely on ChatGPT for their academic work. The study found that students who used ChatGPT reported more procrastination and a drop in GPA. The negative effects of excessive ChatGPT usage highlight the need for caution in its role in education.

OpenAI gave directors early access to its Sora text-to-video generator. Results ranged from impressive to horrifying. The public release of Sora is still uncertain.

Google has started testing its AI-powered Search Generative Experience (SGE) product on users who didn't opt in. SGE is a language model-powered chatbot that provides organized, paraphrased answers to queries. However, it has been criticized for flawed sourcing, incorrect information, and problematic takes. Despite these issues, Google's push for AI-infused search seems to be continuing. SGE is being tested on complex queries and aims to provide helpful information from a range of web pages. Users may start seeing AI-generated search responses in the near future.

AI models accurately predict beer ratings and suggest compounds to improve taste. This could help save time and money in product development. The models outperformed human experts in predicting beer ratings. Specific compounds were identified that contribute to consumer appreciation of beer. The research could also benefit alcohol-free beers and explore food texture and nutrition.

The tech industry is struggling to define "open-source AI," leading to confusion and potential misuse of the term. The Open Source Initiative (OSI) is working on a definition, but the complexity of AI models and the lack of agreement on criteria pose challenges. Companies like Meta, OpenAI, and Google have differing approaches to openness, with some releasing pretrained models but not the training data. The debate revolves around access to data, restrictions on use, and the need for transparency. A clear definition is crucial for fostering innovation, competition, and user freedoms in the AI community.

Meta pursues AI talent with quick offers, emails from Zuckerberg. The company extends job offers without interviews and increases compensation for employees. CEO Mark Zuckerberg personally recruits researchers from Google's DeepMind unit.

This document is the official code repository for the paper "Usable XAI: 10 Strategies Towards Exploiting Explainability in the LLM Era." It includes implementations of various explanation methods for language model families, such as GPT-2 and LLaMA. The repository also contains case studies, including hallucination detection, LLM response verification, training sample influence estimation, and an analysis of Chain-of-Thoughts (CoTs) in explaining model predictions. The document also mentions future improvements and plans for releasing the explanation methods as a package.

Large language models' emergent abilities may not be as unpredictable as previously thought, according to a new paper by researchers at Stanford University. They argue that the sudden appearance of these abilities is a consequence of how researchers measure the models' performance. The study highlights the need for further research and understanding of large language models' behavior.

Tech giants are aggressively hiring lawyers with expertise in artificial intelligence (AI), according to a report. The demand for AI-focused lawyers is increasing as companies seek legal professionals who can navigate the complex legal and ethical issues surrounding AI technology.

Scientists have created AI models that can communicate and teach each other tasks based on written instructions. The AI models used natural language processing (NLP) to describe what they learned to each other. This breakthrough could lead to advancements in artificial intelligence and help us understand how human brains work.

Utah's new Artificial Intelligence Policy Act (AIPA) brings generative AI under consumer protection laws. It establishes liability for AI violations, requires disclosure in regulated occupations, creates an Office of Artificial Intelligence Policy, and imposes fines for violations. The AIPA aims to protect consumers and will take effect on May 1. URL [https://le.utah.gov/~2024/bills/static/SB0149.html]

Legal departments aim to shed the "cost center" label but inadvertently reinforce it, according to Corporate Counsel. The article discusses the perception of legal departments as cost centers and the challenges they face in changing this perception. Read more at: [Article Link]

Researchers have released a large AI training dataset composed of public domain text, showing that language models can be trained without copyrighted materials. Fairly Trained has certified its first large language model, KL3M, developed by 273 Ventures using a curated dataset of legal, financial, and regulatory documents. Common Corpus, the largest available AI dataset for language models, has been released, offering researchers access to a vetted training set free from copyright concerns. These developments demonstrate a growing trend towards legally and ethically created large language models.

Progress in AI is slowing down as the field enters a "grind phase" of refining existing technology. While initial breakthroughs generated excitement, recent advancements have been limited, with little improvement in chatbots and image generators. The current approach lacks true understanding and perception. The author argues that the slowdown is a natural part of progress and should be expected. Creating a truly thinking machine is still decades away.

AI chatbot hype is facing a reality check as generative AI falls short of expectations. Issues such as embarrassing errors, intellectual property infringement, and environmental impact have dampened enthusiasm. Startups in the field are struggling, and companies are finding it difficult to scale generative AI. However, despite the setbacks, there is still optimism for the future of generative AI, with the expectation of improved models and refined data sets. Educators are divided on the impact of AI on their careers. The Axios What's Next Summit discussed the dominance of AI and air travel, as well as the challenges of partisan disagreements. TikTok's cash-for-creators program could lead to the spread of misinformation on the platform.

AI 'ghosts' could harm mental health by interfering with the grieving process, causing confusion, stress, depression, and paranoia. Resurrecting deceased loved ones as avatars may offer comfort, but research suggests it should be used temporarily to avoid emotional dependence. There are also risks of AI ghost-bots saying harmful things or giving bad advice. Forgetting is important for healthy grief, and new ways of remembering the deceased should be found. Oversight and human supervision are necessary for this technology. (Word count: 68)

FTC is investigating Reddit's data-licensing practices ahead of its IPO. Reddit plans to sell shares for $31 to $34 apiece, valuing the company at around $6.5 billion. The investigation focuses on the sale, licensing, or sharing of user-generated content with third parties for training AI models.

In-house attorneys continue to embrace ESG (environmental, social, and governance) but face challenges in staying updated, according to Corporate Counsel. The article explores the struggle of attorneys to keep up with ESG trends and highlights the importance of staying informed in this area.

Apple researchers achieve breakthroughs in multimodal AI, enabling more powerful and flexible AI systems. The research demonstrates the importance of combining different types of training data and model architectures. Apple is ramping up its investments in AI and plans to integrate generative AI capabilities into its products. The company is working on a large language model framework called "Ajax" and a chatbot known as "Apple GPT." Apple's goal is to use AI in various applications, such as personalized playlists, code writing assistance, and open-ended conversation. The MM1 research shows Apple's potential in AI, but the company faces competition in the AI arms race.

Recent research raises concerns about consumer-facing legal AI tools, like ChatGPT, being used for legal information and advice. Risks are high for litigants in lower courts, those with incorrect premises, or unable to assess reliability of responses. Despite growing consumer appetite, if the information provided is wrong, it can be detrimental. The private sector's increased role in delivering digital legal information and advice raises questions about protecting the public without existing legal services regulation. The Law Society is considering options for responding to these developments and calls for urgent action to protect consumers.

Florida attorney Thomas Grant Neusom has been suspended from practicing in the U.S. District Court for the Middle District of Florida for one year. This comes after he filed cases that were found to be "completely fabricated." Neusom violated court rules and Florida's Rules of Professional Conduct by submitting pleadings with frivolous legal arguments based on fabricated cases. The disciplinary committee found that Neusom failed to exercise reasonable due diligence and that some of his legal authorities were completely fabricated.

Judges defend AI use by attorneys for early brief drafts amid concerns about ChatGPT. Some judges support the use of generative AI as a drafting tool, emphasizing the need to verify accuracy. Courts are addressing the impact of AI on legal practice and considering rules for lawyers to certify the accuracy of AI-generated materials. Attorneys using AI must verify information and avoid inputting confidential data into free AI services.

Morgan Stanley promotes Jeff McMillan to head of firm-wide artificial intelligence, becoming the first major Wall Street firm to create a solution based on OpenAI's GPT-4. The move reflects the rising importance of AI in financial services and the competition among firms to fill AI positions.

Breakthrough AI can predict Alzheimer's up to 7 years in advance by analyzing health records and identifying patterns. The AI system accurately predicted Alzheimer's development 72% of the time. The analysis considered various risk factors such as high blood pressure, high cholesterol, vitamin D deficiency, depression, and other conditions. The findings provide insights into the disease's causes and potential vulnerabilities. The research was conducted by the University of California, San Francisco (UCSF) and Stanford University. Article Link [https://www.sciencealert.com/breakthrough-ai-can-now-predict-alzheimers-up-to-7-years-in-advance]

AI agents like Devin, developed by Cognition AI, can perform tasks usually done by software engineers, generating code, planning solutions, and implementing them. These agents are part of a trend where AI can take action to solve problems instead of just providing advice. While there is potential for error, startups like Cognition AI and Google DeepMind are developing AI agents for various tasks, from software engineering to playing video games. The future holds more advancements in AI agents, combining large language models with training in specific domains to create more capable and reliable agents.

Researchers have discovered an attack that can decipher AI assistant responses, exposing the content of private conversations. The attack exploits a side channel present in major AI assistants, allowing an adversary to infer the specific topic of captured responses. The attack has been successful in decrypting responses from OpenAI's ChatGPT and Microsoft's Copilot. The researchers propose mitigation strategies such as sending tokens in large batches or applying padding to packets. OpenAI and Cloudflare have implemented padding mitigations, while Microsoft plans to address the issue with an update.

Extractive AI takes a comprehensive and transparent approach to machine intelligence. It derives meaning and intent using layered reference ontologies, resulting in a more descriptive enumeration of the problem space. Extractive AI provides more structured, consistent, and precise responses, with traceable crumb trails back to the sources. It can assign confidence intervals and adjust recommendations based on the decision environment. While generative AI is exciting, extractive AI offers precise answers with serious implications. Both types of AI have their place and can be utilized for different purposes.

The European Parliament has approved the EU AI Act, a landmark in AI regulation. The act creates tiers of risks and requirements for AI technologies, including human oversight and data governance. It takes effect 20 days after publication. Prohibitions on high-risk AI systems will apply after two years. The European AI Office will be established to govern the act. Stakeholders can get involved early and the act can still be improved. The approval was celebrated, but there were disagreements, such as the use of biometric cameras. The act aims to strike a balance between protecting human rights and promoting AI innovation.

Strategy & Artificial Intelligence. Item ID: 65ef12dee1dc0e7e676435cc. Created: March 11, 2024. Last edited: March 12, 2024. Top 10: No. URL: [link]. Status: Active.

Major technology firms are tempering expectations about generative AI, saying the hype has surpassed its actual capabilities and value for customers. Executives and salespeople at cloud providers like Microsoft, Amazon Web Services, and Google reveal that customers are cautious about increasing spending on AI services due to high costs, accuracy issues, and uncertainty about the value they will receive.

Outsourcing cognitive tasks to technology can lead to skill erosion, where individuals lose the ability to perform those tasks themselves. This can have negative consequences, as shown by examples such as plane crashes and incorrect GPS directions. Skill erosion can cripple organizations, as seen in an accounting company that had to relearn fixed-asset accounting skills after relying too heavily on software. To prevent skill erosion, individuals should pay attention to what automated systems are doing, keep their competence up to date, and critically assess the results. AI presents opportunities but also carries the risk of skill erosion.

This blog post discusses jailbreaking LLMs (Language Learning Models) using ASCII art. It mentions that talking to ChatGPT in Morse code or base64 can bypass restrictions. The post provides links to articles by Eelko de Vos demonstrating jailbreaking ChatGPT using Morse code and base64.

Cognition AI has developed Devin, a software development assistant that can complete entire software projects on its own. The company has raised $21 million in funding and is backed by Peter Thiel's venture capital firm. Devin is described as a next-level assistant that can perform tasks based on natural language commands and can fix bugs on its own. The founders of Cognition AI have a background in sport-coding and have won multiple gold medals in international coding competitions. The company claims to have made a breakthrough in a computer's ability to reason, setting it apart from other coding assistants. Cognition AI is seen as a promising player in the AI coding space, alongside other companies like Magic AI. The impact of coding AIs on the software industry and job market is still uncertain.

Large language models like ChatGPT absorb racist views from the internet. Efforts to make them less toxic curb overt racism but let covert stereotypes grow stronger. Feedback training helps combat overt stereotypes but fails on covert biases. Covert stereotypes strengthen as models increase in size. Companies need better tools to fight racial bias. AI models are used in critical decision-making, such as asylum cases and crime prediction. Discrimination can occur in hiring and screening applications based on race and gender.

Elon Musk's AI startup xAI will open-source its ChatGPT challenger "Grok" after suing OpenAI for abandoning its original mission. The move aligns xAI with other firms like Meta and Mistral that have open-source AI models. Musk has been advocating for open-source AI and wants to establish a third-party referee to oversee AI development. However, concerns have been raised about the potential misuse of open-source AI models. Musk launched xAI to create a truth-seeking AI alternative to OpenAI and Google. Grok was rolled out for Premium+ subscribers of X.

New DOJ Whistleblower Program creates concerns for companies. Program details and impact discussed in article. URL: [link]

Queen's Law is leading the way in reforming legal education by offering a one-year practicum course called Conflict Analytics Lab (CAL) that focuses on tech projects using artificial intelligence. CAL has developed an open source legal AI tool called OpenJustice, which is available for free to all Canadian law schools. The program aims to train students in building, using, and evaluating technology, and encourages interdisciplinary collaboration between law, business, and computer science students. The goal is to increase accessibility to legal services and improve legal technology. Law students are learning foundational skills in generative AI and prompt engineering, preparing them for the future of the legal profession.

AI poses risks to quality of life, institutions, and mental well-being. It promotes mistrust, undermines credibility, exacerbates inequity and bias, and impacts industries. AI could worsen existing risks and harms. Panelists disagree on whether AI poses an existential risk. Policies include transparency, oversight, and diversity. Research should focus on AI's impact on society and building resilient communities.

Baidu's ERNIE Bot surpasses OpenAI's GPT-4 in Chinese. Baidu's CEO foresees a future where only English and Chinese remain as programming languages. AI development has been slower than expected. AI will eliminate the need for programming positions. Li suggests that duplication in investments in large language models is wasteful.

Scientists at The Universities of Manchester and Oxford have developed an AI framework called CLASSIX that can identify and track new and concerning COVID-19 variants. The framework combines dimension reduction techniques and machine learning to quickly identify groups of viral genomes that might present a risk in the future. This method allows automation of tasks that currently require extensive manual curation, reducing resource needs and enabling more researchers to identify concerning pathogen strains. The study was presented in the journal PNAS. [DOI: 10.1073/pnas.2317284121]

Silicon Valley's high cost of working with AI models is pushing academics out of the field. Tech giants like Meta, Google, and Microsoft are funneling billions into AI, creating a resources gap with universities. Scholars frequently partner with tech employees for computing power and data. Big tech companies dominate breakthroughs, shaping the field and pushing researchers to tailor their work for commercial use. Fei-Fei Li, a former Google employee and Stanford professor, is advocating for funding a national AI repository. Policymakers are taking steps to address funding gaps, but scholars say it may not be enough.

The concept of "AI safety" is expanding to include various concerns, from human extinction to algorithmic bias. However, some argue that lumping all these concerns together is unwise, as different risks should be addressed differently. There is a concern that including political concepts like content moderation with non-political issues may politicize AI safety. The term "AI safety" is becoming an umbrella term that encompasses almost every potential downside of software automation. Some advocate for a broader definition of AI safety, while others argue for a more narrow focus.

A pilot program by Walmart and RadNet brings 3D mammography screenings to Walmart Supercenters, increasing access to mammograms. The technology offers more precise and detailed views of breast tissue, allowing for earlier and more accurate detection of breast cancer. The convenience and accuracy of these screenings could bridge the gap in early detection, leading to less invasive treatments and higher survival rates. The project aims to improve prospects and reduce the emotional toll of breast cancer.

Guy Chambers, songwriter for Robbie Williams and Kylie Minogue, expresses concern about the increasing use of AI in the music industry. He warns that AI has the potential to replace songwriters and suggests that future albums may need disclaimers about their human involvement. Chambers explains how AI can generate music and lyrics based on specific requests, and he acknowledges the benefits of AI but also highlights the dangers and urges young musicians not to become lazy. The use of AI in music creation is a topic of conflict in the industry, with organizations advocating for limitations on AI and emphasizing the unique creativity of human artists.

AI startups face unique challenges and cannot rely on traditional startup strategies. Incumbents are embracing AI, have access to data, and attract top talent. The AI market is not a separate market, but a tool for existing markets. Startups need to differentiate themselves and create distribution quickly. A different strategy is needed to succeed in the AI space.

AI boom in healthcare faces a conundrum: Americans oppose AI replacing doctors but support its use for administrative tasks. With limited access to primary care physicians and high physician burnout rates, AI may be necessary to meet healthcare demands. Generational differences exist, with younger generations expecting AI and other tools in healthcare. The aging population increases the complexity of diseases and exacerbates staff shortages. The healthcare AI industry must address public trust issues.

Head-spinning changes to antitrust enforcement are forcing firms to recalibrate deal-making, impacting corporate counsel. For more details, refer to the article at the provided URL.

AI-designed proteins' potential weaponization discussed in safety guidelines for scientists. (70 tokens)

Companies find it difficult to adopt generative AI tools like LLMs. Suggestions for onboarding include choosing performance over novelty, combining GenAI with vector databases, involving humans in the loop, tracing data, and having realistic expectations. The fear of missing out and the transformative power of GenAI have driven businesses and individuals to explore these technologies. Media articles have highlighted the potential job losses due to GenAI.

Tech is driving the growth of the mid-market, as discussed in the article "How Tech Is Helping Fuel the Rise of the Mid-Market" on Law.com [http://law.com/]. The article explores the impact of technology on the mid-market and provides insights into its role in driving growth. For more details, visit the URL: https://www.law.com/2024/03/08/how-tech-is-helping-fuel-the-rise-of-the-mid-market/ [https://www.law.com/2024/03/08/how-tech-is-helping-fuel-the-rise-of-the-mid-market/].

Missouri's rules regarding the use of artificial intelligence (AI) in court may hinder access to justice for self-represented litigants. A court rule in Missouri exclusively bans self-represented litigants from using generative AI to draft filings, and a recent court decision fined a nonlawyer litigant for using AI-generated cases. These restrictions and sanctions could have a chilling effect on access to justice initiatives involving AI. Instead, judges should consider providing educational resources to pro se litigants about the limitations of AI and publicly available language models. AI has the potential to increase access to justice and legal information.