6. Research
AI Supercharges Coders: Studies reveal AI tools like GitHub Copilot boost programmer productivity by 26%, with junior developers benefiting most from AI-assisted coding.
Details
Several new studies have been investigating the impact of genAI on programming. An analysis of three randomized controlled trials involving 4,867 developers at Microsoft, Accenture, and a Fortune 100 electronics manufacturing company found that those with access to GitHub Copilot—completed 26.08% more tasks than those without the tool. Similarly, a working paper published by the Bank for International Settlements reported on a field experiment using CodeFuse, a large language model developed by Ant Group to assist programming teams. Findings highlighted a significant increase in productivity, measured by the number of lines of code produced, which was most pronounced among junior staff as a result of their more active engagement with the tool. Source: SSRN, Source: Bank for International Settlements
AI vs. Conspiracies: An AI chatbot powered by GPT-4 successfully debunks conspiracy theories, reducing believers’ conviction by 21% and sustaining the change over time.
Details
Researchers have developed an AI chatbot using OpenAI’s GPT-4-Turbo that can debunk conspiracy theories and influence people to reconsider their beliefs. The study involved participants interacting with the chatbot, which provided detailed counterarguments to their conspiracy theories. The researchers recruited over 1,000 participants, who were asked to describe a conspiracy theory they believed in and rate their conviction. After interacting with the chatbot, participants’ confidence in their chosen conspiracy theory decreased by an average of 21%. A follow-up survey two months later showed that the shift in perspective had persisted for many participants. Source: Nature.com
AI in Finance: Financial regulators in New Zealand, Canada, and Sweden explore AI adoption in finance, with studies showing proactive integration and positive reception to AI-driven services.
Details
New Zealand’s Financial Markets Authority (FMA) has published new research on the use of AI in financial services, revealing a cautious yet proactive approach among firms towards integrating AI technologies. Based on responses from 13 financial service providers across asset management, banking, financial advice, and insurance, the research indicates that all participants are either currently using AI or plan to adopt it soon, motivated by goals such as enhancing customer outcomes, improving operational efficiency, and strengthening fraud detection and risk management. Concurrently, Canada’s Ontario Securities Commission (OSC) has published a report on how AI influences retail investor decision-making, based on a behavioural science experiment involving an online investment simulation where participants were given a hypothetical $20,000 to invest and received suggestions from either a human financial services provider, an AI tool, or a combination of both. Findings indicate the Canadian are receptive to AI-driven financial advice. Meanwhile, the Swedish Financial Services Authority (FI) has initiated a study to map out AI utilization in Sweden’s financial sector to deepen the understanding current and anticipated AI use cases, firms’ perceptions of related benefits and challenges as well as the expected impact of the EU AI Act on their operations. Source: New Zealand Financial Markets Authority, Source: Ontario Securities Commission, Source: Swedish Financial Services Authority
7. Dev Alerts
Don’t miss these deprecation and shutdown dates
-
As of September 13, gpt-3.5-turbo-0613 and gpt-3.5-turbo-16k-0613 were officially shutdown. Other model snapchats including gpt-3.5-turbo-0125, gpt-3.5-turbo-1106 remain available.
-
New fine-tuning training runs on babbage-002 and davinci-002 will no longer be supported from October 28, 2024. and developers are recommended to switch to gpt-4o-mini. Existing fine-tunes babbage-002 and davinci-002 will remain accessible.
-
Both gpt-4-vision-preview and gpt-4-1106-vision-preview will be shut down on December 6, 2024. Recommended replacement is gpt-4o.
Explore o1 use cases with two new OpenAI cookbook additions
-
Using reasoning for data validation: The example covers how to use o1-preview to perform data validation on a synthetic medical dataset, demonstrating the process of generating data with intentional errors, validating its accuracy using the model, and analyzing the results to assess the model’s precision and recall in identifying and explaining data issues. Link
-
Using reasoning for routine generation: This example covers the process of using 01-preview for converting customer service knowledge base articles into structured, executable routines for a language model to automate customer interactions efficiently. Link
Find your OpenAI dream role
OpenAI currently has 187 open positions across San Francisco and several other key locations including New York City, London, Tokyo, Dublin, Seattle, and Singapore. Available roles span across all areas with major concentrations in Applied AI, Research, and Go-To-Market.
Get all the insights here
- Global Reach and Diverse Locations:
-
International Expansion: Positions are available in major cities like San Francisco, New York City, London, Tokyo, Dublin, Seattle, and **Singapore highlight OpenAI’s drive to expand its international footprint.
-
Remote Opportunities: Several roles are open to remote candidates, reflecting OpenAI’s flexible work culture and adaptation to the evolving workplace landscape.
- Dominance of Engineering and Research Roles:
-
Engineering Focus: A significant portion of the openings are in Applied AI Engineering, Platform Engineering, Research, and Security.
-
Specialized Positions: Several highly specialized roles are available like GPU Kernels Engineer, HW/SW Co-Design Engineer, and Distributed Training Engineer.
- Strengthening Business and Customer Relations:
-
Sales and Account Management: Positions such as Account Director, Sales Leader, and Customer Success Manager indicate efforts to expand client engagement and market reach.
-
Global Affairs and Partnerships: Roles like Japan Policy & Partnerships Lead and Senior Manager, Partner Management emphasize building strategic alliances and navigating regional regulations.
- Investment in Human-Centric AI Development:
-
Human Data and Alignment: Jobs in Human Data and Alignment departments suggest a focus on refining AI models to be more aligned with human values and ethics.
-
Safety and Security: Positions like Research Scientist, Model Safety and Security Engineer, Detection & Response highlight a commitment to developing safe and secure AI systems.
- Operational and Organizational Growth:
-
People and Workplace Management: Roles such as Benefits Operations Specialist and Workplace Operations Manager reflect efforts to enhance employee experience and operational efficiency.
-
Finance and Legal: Openings for Revenue Accounting Manager, Tax Director, and Head of Internal Controls indicate a scaling of financial operations and compliance functions.
- Enhancing Communication and Public Engagement:
-
Communications and Marketing: Positions in Media Relations, Digital Marketing Lead, and Events Operations Manager show a drive to amplify OpenAI’s public presence and stakeholder engagement.
-
Design and User Experience: Roles like Design Engineer, Communications Design and Front End Software Engineer focus on improving user interfaces and overall experience.