Subscribe & Follow
Jobs
- Sales and Business Development Manager Cape Town
- Content Curator Ilovo, Sandton
- Digital Archive Intern Cape Town
- Junior Digital Art and Social Media Marketing Coordinator Johannesburg
- Digital Marketing Learnership Cape Town
- CRM Digital Manager (6-Month Contract) Johannesburg
- CRM Digital Admin / Consultant (6-Month Contract) Johannesburg
- Social Media Manager Cape Town
- Videographer and Video Editor Tshwane
- Multi Media Journalist | South Coast Sun Durban
#MDPMI: Open AI's ChatGPT surprises founders with mass adoption
Excitement
"In November 2022, we released ChatGPT as a free research preview. We were not expecting the level of excitement and interest that people showed when they used it. The technology behind it had been publicly available for some time, so the extent of its adoption was beyond our expectations and honestly surprised us and many others," shared Haidee Schwartz, senior counsel for Open AI.
Schwartz made these remarks during the Media and Digital Platforms Market Inquiry (MDPMI) on Monday, as the public hearings commenced with over 40 key stakeholders. The discussions centred on AI technology, where Open AI's advancements were highlighted, including the robotic hand Dactyl, the coding assistant Codex, and the DALL EA neural network model, capable of generating images from textual descriptions.
This revelation came as Open AI, primarily focused on AI research, found itself with an unexpected consumer-facing hit. However, with the widespread use of AI chatbots like ChatGPT, a host of concerns have emerged, touching on security, data leakage, confidentiality, liability, intellectual property, compliance, and privacy.
Public information
Addressing the concerns surrounding data usage, Nora Puckett, general counsel for Open AI, emphasised their commitment to data security. "After training, the model does not retain access to the training data. It does not function like a search engine," she stated.
Developing an advanced language model involves a process called pre-training, which requires exposing the model to vast amounts of human knowledge over months. Post-training, Open AI further refines the technology for safety and utility.
Puckett clarified that Open AI's platforms rely solely on publicly available information, excluding data from paywalls or the 'dark web.' They have also established agreements with certain organisations, such as media houses, to use their information for AI training. However, website owners with publicly available data can choose to opt-out of Open AI's data usage.
Commissioner James Hodge said more hearings into organisations such as Google will occur this week.