AI-generated summary of CPD session: Generative AI: what can we expect in the year ahead?
Zoom has its own AI Companion which offers a Quick recap, Next steps and a Summary. Below is the unedited summary of the session before any remediation. And below that are some recommended resources that were mentioned by some of the attendees during the session.
Rapid Evolution of Large Language Models
Paulo discussed the rapid evolution of AI, particularly large language models (LLMs) such as Gemini, Chat GPT, and Claude. He explained that these models are based on a large bank of data and use sophisticated algorithms to predict the likelihood of text based on user prompts. Paulo also mentioned that OpenAI's decision to release GPT-3 before it was ready led to its rapid adoption, reaching 100 million users in just 10 days. He noted that since then, other models like Claude and Gemini have been launched, leading to a competitive landscape in the field.
Large Language Models and DeepSeek
Paulo discussed the advancements in large language models, emphasizing that improvements are not always linear and depend on the specific task. He mentioned that GPT-4 and GPT-3 are good at handling different types of data, and the choice of model depends on the user's preference and task. Paulo also introduced DeepSeek, a model that achieved similar quality to its American counterparts at a lower cost, using simpler and fewer chipsets and less energy. He suggested that users experiment with different models to find the best fit for their needs.
Deep Seek AI Model Discussion
Paulo discussed the workings of Deep Seek, an open-source, Chinese-owned AI model. He noted that while it's difficult to predict government influence on such companies, Deep Seek's open-source nature and potential for government influence could lead to censorship. Paulo also highlighted the cost-effectiveness of Deep Seek, stating that it's 240 times cheaper than similar models two years ago. He emphasized the importance of understanding how data is processed and stored when using AI models like Deep Seek, and suggested that different models might be used for different tasks depending on the confidentiality agreements in place.
AI Usage in Google Workspace
Paulo discussed the increasing use of AI in various tools, including Google Workspace and Microsoft services. He mentioned that AI is likely being used in the background, even if users are not aware of it. Paulo also addressed concerns about AI usage in publishing houses, noting that some companies are allowing editors to set up internal AI bots for specific tasks. He mentioned that the cost of running these personalized bots has decreased significantly, making them more accessible to smaller companies. Paulo hinted at the possibility of running another AI course later in the year and encouraged those interested to join.
Large Language Models and Copyright
Paulo discussed the use of large language models in their company's services, specifically for private teachers to design personalized lesson plans. He explained that the cost of running this system is roughly half a cent per lesson plan. Paulo also addressed potential copyright issues when uploading course book pages for analysis, stating that as long as the content is not stored or used to train the AI model, it should not be a problem. He mentioned that most of the features they use require a paid subscription, costing around $20 a month. Paulo also touched on the possibility of sharing materials through a personalized bot, which would remain private.
Generative AI in Editorial Tasks
Paulo discussed the increasing use of generative AI in editorial tasks, noting that almost 50% of editors are regularly using it. He highlighted that AI is becoming a requirement for several editorial jobs and that companies are developing guidelines for its fair and responsible use. Paulo also mentioned that the AI market in education is projected to reach $12 billion by 2027, and that publishing houses are trialing or licensing large language models for editorial tasks. He shared his personal experience of seeing an influence of AI in the type of work he gets commissioned, and predicted that the use of AI in editorial tasks will become more widespread in the next few years.
Detecting AI-Generated Content Challenges
Paulo discussed the challenges of detecting AI-generated content, particularly when using a knowledge base not publicly available. He suggested that AI-generated content is difficult to identify due to its similarity to human-generated content. Paulo also mentioned the use of AI in the initial stages of writing, which can save up to 30% of time. He noted that AI is being increasingly used for copy editing and proofreading, and questioned the validity of the "no AI" rule, as it's unclear what would be considered AI. He concluded by stating that AI is becoming more prevalent and accepted in the writing process.
AI's Impact on Translation and Jobs
Paulo discussed the increasing use of AI in translation, noting that the average quality of non-literary translation by AI is often better than that of an inexperienced translator. He suggested that translators may become quality checkers rather than translators themselves. Paulo also highlighted the ongoing legal battle over copyright and the potential for AI to reduce project rates and times. He concluded by mentioning a report from LinkedIn that AI is not replacing jobs, but rather creating new ones for those who understand how to effectively use AI.
AI's Role in Future Work Environments
Paulo discussed the future of work in the context of AI and automation. He acknowledged the potential for AI to replace certain tasks, such as summarizing and initial searches, but emphasized the need for humans to manipulate these tools to work faster. Paulo also predicted that AI will become more prevalent in publishing houses, with some already piloting AI projects. He suggested that while AI can perform proofreading, content checking and quality assurance should still be done by humans. Paulo also noted that AI-generated content lacks the creative and human elements that humans bring to writing. He concluded by suggesting that the future of work will involve more interaction with AI tools, with humans acting as prompters, editors, and polishers until AI can replace these roles.
AI in Publishing and Content Creation
Paulo discussed the rapid development and integration of AI in various industries, particularly in the publishing sector. He highlighted the emergence of AI tools like Sora, Vo2, and MakeaVideo, and their potential to revolutionize content creation. Paulo also touched on the challenges of AI-generated content, such as quality and originality, and the potential for AI to replicate roles traditionally held by publishers. Karen announced the upcoming courses on AI for freelancers and publishers, and encouraged attendees to sign up for updates. The conversation ended with Paulo sharing a video entirely produced by AI, demonstrating the current state of AI-generated content.
AI-generated content may be inaccurate or misleading. Always check for accuracy.
Recommended resources from our attendees
- https://www.cathoven.com/ (for question generating)
- https://www.theguardian.com/technology/2025/jan/28/we-tried-out-deepseek-it-works-well-until-we-asked-it-about-tiananmen-square-and-taiwan
- https://www.tuc.org.uk/research-analysis/reports/artificial-intelligence-regulation-and-employment-rights-bill The Trade Union Congress (TUC) have generated a bill which is relevant to employers using AI in the workplace: Artificial Intelligence (Regulation and Employment Rights) Bill)
- https://www.oneusefulthing.org/p/which-ai-to-use-now-an-updated-opinionated?utm_source=share&utm_medium=android&r=2is7x9&triedRedirect=true (Which AI to use now: and updated opinionated guide by Ethan Mollick)