Former OpenAI VP says human taste will matter more in the world where AI is making slop
Even with everything that AI is capable of doing, there's one thing it still can't do properly, i.e., think and feel like a human. That's the point Krithika Shankarraman, former VP of marketing at OpenAI, is making. She believes that in a world flooded with AI-made content, it's the human touch — our ideas, our choices, and our care — that will make the difference. In a recent episode of Lenny's Podcast (via a Business Insider's report), she said, 'Taste is going to become a distinguishing factor in the age of AI because there's going to be so much drivel that is generated by AI. That power is at anyone's fingertips.'advertisementShankarraman warns that while AI can speed up work, it's not meant to replace people. She believes that AI should support us with our work, and not take over. If businesses depend too much on AI and leave humans out of the process, their work will all start to look the same. 'The companies that are going to distinguish themselves are the ones that show their craft,' she said. 'That they show their true understanding of the product, the true understanding of their customer, and connect the two in meaningful ways.' In short, the best companies will be the ones that care about their products and their customers. And only real people can make those connections. No matter how advanced AI becomes, it still can't replace human care and creativity. She added, 'What it means to market a product, what it means to show up as a fantastic operator, is in and of itself changing.'advertisement
To keep up, Shankarraman believes it's important to understand the basics. That's why she supports learning STEM (science, technology, engineering, and maths). According to her, 'This is why I would still be a very firm believer in STEM education,' she said. 'You understand the fundamental concepts. And then you can have a choice and optionality in how you decide to apply those concepts, but the concepts themselves have to be there in the foundations.'Shankarraman also pointed out that learning just to pass exams isn't helpful anymore. We should be learning to understand how things work, so we're better prepared to adapt and grow. 'Because being of that growth mindset, if you go to school just to earn the grades or to finish the coursework, it's a very different mindset than if you go to school to learn those concepts and to understand how to apply them,' she said.Shankarraman said individuals must take responsibility for how they use AI. But she also hopes that companies don't get stuck in a race to show off who has the best chatbot. Instead, she wants them to think long-term and use AI to make a real difference. 'Long story short, what I'm trying to say is that all of these companies have to think in a much more long-term oriented fashion. Because it's not about a race of the best chatbot and the best outputs. It's about, how does AI become a positive force for humanity?', she said.
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Hindustan Times
an hour ago
- Hindustan Times
Thinking AI models collapse in face of complex problems, Apple researchers find
Just days ahead of the much-anticipated Worldwide Developer Conference (WWDC), Apple has released a study titled 'The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity', which saw researchers testing 'reasoning'; AI models such as Anthropic's Claude, OpenAI's o models, DeepSeek R1 and Google's Thinking models to see how far they can scale to replicate human reasoning. Spoiler alert — not as much, as the entire AI marketing pitch, would have you believe. Could this signal what may be in store for Apple's AI conversation ahead of the keynote? The study questions the current standard evaluation of Large Reasoning Models (LRMs) using established mathematical and coding benchmarks, arguing they suffer from data contamination and don't reveal insights into reasoning trace structure and quality. Instead, it proposes a controlled experimental testbed using algorithmic puzzle environments. The limitations of AI benchmarking, and need to evolve, is something we had written about earlier. 'We show that state-of-the-art LRMs (e.g., o3-mini, DeepSeek-R1, Claude-3.7-Sonnet-Thinking) still fail to develop generalizable problem-solving capabilities, with accuracy ultimately collapsing to zero beyond certain complexities across different environments,' the researcher paper points out. These findings are a stark warning to the industry — current LLMs are far from general-purpose reasoners. The emergence of Large Reasoning Models (LRMs), such as OpenAI's o1/o3, DeepSeek-R1, Claude 3.7 Sonnet Thinking, and Gemini Thinking, has been hailed as a significant advancement, potentially marking steps toward more general artificial intelligence. These models characteristically generate responses following detailed 'thinking processes', such as a long Chain-of-Thought sequence, before providing a final answer. While they have shown promising results on various reasoning benchmarks, the capability of benchmarks to judge rapidly evolving models, itself is in doubt. The researchers cite a comparison between non-thinking LLMs and their 'thinking' evolution. 'At low complexity, non-thinking models are more accurate and token-efficient. As complexity increases, reasoning models outperform but require more tokens—until both collapse beyond a critical threshold, with shorter traces,' they say. The illustrative example of the Claude 3.7 Sonnet and Claude 3.7 Sonnet Thinking illustrates how both models retain accuracy till complexity level three, after which the standard LLM sees a significant drop, something the thinking model too suffers from, a couple of levels later. At the same time, the thinking model is using significantly more tokens. This research attempted to challenge prevailing evaluation paradigms, which often rely on established mathematical and coding benchmarks, which are otherwise susceptible to data contamination. Such benchmarks also primarily focus on final answer accuracy, providing limited insight into the reasoning process itself, something that is the key differentiator for a 'thinking' model compared with a simpler large language model. To address these gaps, the study utilises controllable puzzle environments — Tower of Hanoi, Checker Jumping, River Crossing, and Blocks World — and these puzzles allow for precise manipulation of problem complexity while maintaining consistent logical structures and rules that must be explicitly followed. That structure theoretically opens a window, a glance at how these models attempt to 'think'. The findings from this controlled experimental setup reveal significant limitations in current frontier LRMs. One of the most striking observations is the complete accuracy collapse that occurs beyond certain complexity thresholds across all tested reasoning models. This is not a gradual degradation but a sharp drop to near-zero accuracy as problems become sufficiently difficult. 'The state-of-the-art LRMs (e.g., o3-mini, DeepSeek-R1, Claude-3.7-Sonnet-Thinking) still fail to develop generalizable problem-solving capabilities, with accuracy ultimately collapsing to zero beyond certain complexities across different environments,' note the researchers. These results inevitably challenge any notion that the LRMs truly possess generalisation problem-solving skills, required for planning tasks or multi-step processes. The study also identifies a counter-intuitive scaling limit in the models' reasoning effort (this is measured by the inference token usage during the 'thinking' phase), which sees these models initially spend more tokens, but as complexity increases, they actually reduce reasoning effort closer to the inevitable accuracy collapse. Researchers say that 'despite these claims and performance advancements, the fundamental benefits and limitations of LRMs remain insufficiently understood. Critical questions still persist: Are these models capable of generalizable reasoning, or are they leveraging different forms of pattern matching?,' they ask. There are further questions pertaining to performance scaling with increasing problem complexity, comparisons to the non-thinking standard LLM counterparts when provided with the same inference token compute, and around inherent limitations of current reasoning approaches, as well as improvements that might be necessary to advance toward more robust reasoning. Where do we go from here? The researchers make it clear that their test methodology too has limitations. 'While our puzzle environments enable controlled experimentation with fine-grained control over problem complexity, they represent a narrow slice of reasoning tasks and may not capture the diversity of real-world or knowledge intensive reasoning problems,' they say. They do add that the use of 'deterministic puzzle simulators assumes that reasoning can be perfectly validated' at every step, a validation that may not be feasible to such precision in less structured domains. That they say, would restrict validity of analysis to more reasoning. There is little argument that LRMs represent progress, particularly for the relevance of AI. Yet, this study highlights that not all reasoning models are capable of robust, generalisable reasoning, particularly in the face of increasing complexity. These findings, ahead of WWDC 2025, and from Apple's own researchers, may suggest that any AI reasoning announcements will likely be pragmatic. The focus areas could include specific use cases where current AI methodology is reliable (the research paper indicates lower to medium complexity, less reliance on flawless long-sequence execution) and potentially integrating neural models with traditional computing approaches to handle the complexities where LRMs currently fail. The era of Large Reasoning Models is here, but this 'Illusion of thinking' study is that AI with true reasoning, remains a mirage.
Time of India
6 hours ago
- Time of India
Amazon freezes retail hiring budget for 2025 amid job cuts
Amazon is freezing its hiring budget for its retail business this year, according to a report by Business Insider. As per the report, an internal email revealed that the Seattle-based retailer plans to keep "flat headcount opex," or operating expense, this year compared to last. These costs include employee salaries and stock-based compensation. An Amazon executive shared earlier this year that any rise in the hiring budget will face close scrutiny and require strong justifications. The retail business is shifting its focus from headcount targets to managing teams based on fixed operating budgets. Amazon's retail division covers a wide range of operations, including its online store, logistics network, and fresh grocery service. These changes affect only corporate staff in Amazon's retail division, not warehouse workers or those in Amazon Web Services, the company's cloud division. An Amazon spokesperson told Business Insider that the company will continue hiring, and a freeze on increasing the hiring budget doesn't mean recruitment is stopping. "Each of Amazon's many businesses has its own approach to hiring based on its individual needs," said Hoffman. "However, across the company, we've historically considered both the number of people we need to hire and the associated costs — that is, operating expenses or opex — of those hiring decisions." This report immediately follows the job cut announcement in Amazon's books division, including at Goodreads and Kindle units. The company stated that fewer than 100 employees were impacted, and that the aim was to boost efficiency and streamline operations. The news aligns with CEO Andy Jassy's ongoing focus on improving profit margins and operational efficiency. He is also working to cut down on what he describes as excessive bureaucracy within the company, including reducing the number of managerial roles. Last month, Amazon also laid off workers in its devices and services division. However, it is to be noted that, despite these reductions, the company added approximately 4,000 jobs in the first quarter of this year compared to the same period last year.
Hindustan Times
6 hours ago
- Hindustan Times
You can now schedule tasks with Gemini as Google's powerful new AI feature rivals ChatGPT's capabilities
Google is steadily evolving Gemini into a smarter, more proactive AI assistant that now competes directly with OpenAI's ChatGPT. The tech giant has started rolling out a feature called Scheduled Actions, which lets users automate recurring or timed tasks without repeating commands. Originally previewed during Google I/O, Scheduled Actions is now arriving on both Android and iOS devices. The feature is currently available to subscribers of Google One AI Premium and select Google Workspace business and education plans. With this rollout, Google is pushing Gemini closer to becoming a fully integrated productivity companion. Scheduled Actions let users instruct Gemini to perform specific tasks at set times or intervals. This includes sending daily calendar summaries, weekly content prompts, or even one time reminders. Once scheduled, Gemini handles them automatically in the background with no follow up required. For example, a user might say, 'Send me a summary of today's meetings every morning at 8 AM' or 'Generate weekly blog ideas every Friday at 10 AM.' These tasks run quietly behind the scenes, transforming Gemini from a reactive chatbot into a daily-use productivity tool. The setup process is built to be intuitive, making automation easy for both everyday users and professionals. Within the Gemini app, users can define a task, set the time, and choose the frequency through a clean and accessible interface. Scheduled Actions puts Google in direct competition with the kind of automation ChatGPT users create through Zapier or custom workflows. What gives Gemini a clear edge is its deep integration with Google's suite of apps. Functioning across Gmail, Calendar, Docs, and Tasks, Gemini offers a smooth setup and efficient task execution experience. Since it is built into tools people already use, Gemini can interact directly with information across Google's ecosystem. There is no need for third party services or custom scripts. For users already invested in Google's platform, the experience is more seamless than ChatGPT's dependence on external integrations. Scheduled Actions signals a shift in expectations for how AI assistants should function. Instead of waiting for commands, Gemini can now anticipate and handle repetitive tasks, offering a more personal and assistant like experience. While this may be just the beginning, it is a clear step toward positioning Gemini as a truly productivity first AI assistant. And as Gemini continues to evolve, it may not just catch up to ChatGPT but define the next generation of digital assistance.
