Latest news with #ChatGPT-2
Mint
24-05-2025
- Business
- Mint
You've heard of LLMs. Prosus is building its own LCM for commercial smarts.
Dutch technology investment firm Prosus NV is building its own 'large commerce model" to help its portfolio companies do business better, according to a top company executive. 'We're showing our large commerce model how users are traversing our portfolio companies' applications," said Paul van der Boor, vice president of AI at Prosus. He explained it as taking the various touchpoints of a user's journey through an app—from opening the app to raising complaints to getting queries answered—and using all of them to create a model that contains 'commercial intelligence". 'When you have a model like that, that is created on a subset of users' interactions with an app, there are other group companies that can use that model to understand how to better help that customer," said Boor. 'We're trying to learn commercial intelligence based on all the interactions we can piece together. So when a new interaction comes in, we can predict what happens there." Prosus has already begun testing this model with iFood, the Brazilian food delivery startup that was earlier led by Prosus's new chief executive officer Fabricio Bloisi, and OLX, a classified advertising company. Boor expects lessons learnt from implementing the commercial intelligence model in Latin America and Europe will help Prosus' group companies. 'We've already done a couple of tests where we know it expands outside of food. The idea is that, at least today, in iFood, one model exists that can help with a whole range of tasks for their constituents," Boor said. Also read | Prosus to take PE style bets, India one of three focus areas AI in Prosus's portfolio companies Prosus sees the current version of its large commerce model (LCM) as akin to OpenAI's ChatGPT-2, which was released in 2019. 'We know we have a path to get to the GPT-4 level, so we're taking it step by step," Boor said. Going forward, the investment firm expects that both existing and new companies in its portfolio to have an AI-first approach. The shift comes alongside a more focused attitude towards investing as well since Bloisi took charge as CEO of Prosus 10 months ago. Prosus-backed Swiggy Ltd, for example, uses various AI models from the minute a user opens the app as the Bengaluru-headquartered company seeks to make its food-delivery and quick-commerce experience highly personalised—such as preferred foods, restaurants, and delivery time. 'We will encourage those use cases across our whole group of companies. We know it works because of iFood in Brazil, we've seen it in Poland, and we encourage it here in India and the other way around," Boor said. Also read | After lip-smacking Swiggy IPO, Prosus prepares for its next hit Beyond artificial intelligence Alongside its deeper investments in AI, Prosus expects a few other technology aspects to play a big role in its investment decisions. 'Agentic systems that can reason, that can do multiple tasks, and shop with or for you is a big thing. Our thesis is that agents will work for you to help you find what you need," said Boor. 'Second is that we expect all of the user interfaces to change as a result of multi-modality." Multi-modality allows people to use an app through text, voice and even pictures. Prosus expects voice to emerge as a big factor in user experience. Bengaluru-based e-commerce firm Meesho, for example, deploys several AI agents across the customer service workflow, including a voice bot for customer service that currently serves users in Hindi and English. 'We've already seen that in Brazil. It's similar to India because people there are naturally inclined to use AI. It offers a huge opportunity as one of the big factors and how to change the form factor of technology using AI," said Boor. Prosus's notable AI investments this year include participating in a $13.5 million funding round in Luzia, a Spanish AI-based personal assistant app, earlier this month; a $7.25 million seed check to another Latin American personal assistant app called Zapia; and leading a $54 million Series-B funding in Taktile, a decision automation platform. Prosus also closed its $1.7 billion acquisition of Despegar, a Latin American AI-first travel agency, this year. In India, the Dutch firm pumped in nearly $30 million into ride-hailing platform Rapido in February. Last year, Prosus invested $100 million in Mumbai-based Vastu Housing Finance Corp. Ltd and acquired a 10.65% stake for $80 million in supply chain financing startup Mintifi. Also read | Groww to acquire Prosus-backed Fisdom for around $150 million
Yahoo
14-05-2025
- General
- Yahoo
I Thought ChatGPT Was Killing My Students' Skills. It's Killing Something More Important Than That.
This essay was adapted from Phil Christman's newsletter, the Tourist. Subscribe here. Before 2023, my teaching year used to follow a predictable emotional arc. In September, I was always excited, not only about meeting a new crop of first-year writing students but even about the prep work. My lesson-planning sessions would take longer than intended and yet leave me feeling energized. I'd look forward to conference week—the one-on-one meetings I try to hold with every student, every term, at least once—and even to the first stack of papers. In October, predictably, I'd crash a little bit, but by late November, I'd be seeing evidence that even some of my least enthusiastic students were beginning to take in everything I'd been trying to tell them. By the time classes ended, I'd be loving everything about my job again, especially in the years when kids would stay behind on the last day to shake my hand and say thank you, or write me a note. The second semester would go roughly the same way. The exhaustion would hit a little earlier, which made the recovery all the sweeter. The funny thing about this cycle is that I would forget, every time, that it was a cycle. In October, in March, I would genuinely believe that I had never had a group of students who had let me down like this before, and that I had never let myself and a group of students down to this extent before. The crash was new each time. Oh, sure, I thought, a year ago I kind of felt this way, but this time I have solid reasons—last year's solid reasons having already evaporated from my memory. The intensity of teaching brings a certain amnesia with it, like marathoning and—I am told—childbirth. I only know I go through this cycle because my wife watches me go through it every year, and reminds me. She remembers last year's solid reasons even if I don't. Since the 2022–23 school year, when ChatGPT-2 and then -3 hit the scene, this cycle now has a new component. About a week or so after the end-of-semester Good Feelings Era, I read the latest big journalistic exposé about the ubiquity of college-level Chat-GP-Cheat and start wondering whether anyone learned anything. As I end yet another semester, I have my pick of such articles, whether it's this ambivalent longer view from the New Yorker or this rather sensational on-the-ground exposé from New York magazine. The latter article begins by introducing us to a student named Lee (not his real name): Lee was born in South Korea and grew up outside Atlanta, where his parents run a college-prep consulting business. … When he started at Columbia as a sophomore this past September, he didn't worry much about academics or his GPA. 'Most assignments in college are not relevant,' he told me. 'They're hackable by AI, and I just had no interest in doing them.' While other new students fretted over the university's rigorous core curriculum … Lee used AI to breeze through with minimal effort. When I asked him why he had gone through so much trouble to get to an Ivy League university only to off-load all of the learning to a robot, he said, 'It's the best place to meet your co-founder and your wife.' 'The best place to meet your co-founder and your wife'! Only well-off people expect life to be this predictable; everybody else knows better. In fact, there are, if we have eyes to see them, many reasons in this early paragraph not to panic, not to feel that anything has fundamentally changed. Lee's parents, we're told, run a test-prep company, which means that he's part of a large, existing system that already treats education as a series of robotic steps even as it pretends to value students and learning. Well, any longtime writing instructor knows that there's no real way to stop a determined rich kid from cheating their way through a writing class. If nothing else, they can always afford to pay someone to write a paper for them—and even if you think you've attuned your paper requirements so thoroughly to your assigned readings and class discussions that a bought paper will fail your rubric, they can probably always pay someone enough to fake that. For ye have the rich always with you. Lee is almost charmingly brazen in his lack of integrity, almost innocent in his seeming ignorance of the possibility of having it. After he gets hauled into Columbia's honor court because he built a business helping other kids cheat their way through remote interviews, his story concludes thus: 'Lee thought it absurd that Columbia, which had a partnership with ChatGPT's parent company, OpenAI, would punish him for innovating with AI.' There's hope for Lee yet. Though maybe not for Columbia University, governed as it is by people who aren't even capable of this insight. Lee is a problem, but a problem of a sort that I'm familiar with. It's a student like Wendy who makes me panic: I asked Wendy if I could read the paper she turned in, and when I opened the document, I was surprised to see the topic: critical pedagogy, the philosophy of education pioneered by Paulo Freire. The philosophy examines the influence of social and political forces on learning and classroom dynamics. Her opening line: 'To what extent is schooling hindering students' cognitive ability to think critically?' Later, I asked Wendy if she recognized the irony in using AI to write not just a paper on critical pedagogy but one that argues learning is what 'makes us truly human.' She wasn't sure what to make of the question. What most worries me about this anecdote—which is perfect in its thematically fractal quality, with the first sentence of Wendy's ersatz essay embodying the intellectual decline that 'her' essay ostensibly describes—is that I'd be reasonably happy if a first-year student turned in something like this. It doesn't have that ChatGPT stiffness that I've learned to look for, and unlike a lot of (fake and real) essays that I always end up tearing apart in the comments, it immediately zeroes in on a single point, rather than messing about with the three-pronged '[Writer] does [X] by doing [thing, thing, and thing]' format that Advanced Placement classes have cursed us with, and that I spend weeks deprogramming my students out of. I would maybe cut 'cognitive' out of the sentence, but it's otherwise unobjectionable. If this is what cheating now looks like, I not only don't know how I'm supposed to tell if my students are cheating—I don't even know how I can be sure they wrote the thank-yous that mean so much to me. ChatGPT, in giving my students an alternative to skill-building, hurts their ability to learn, but more than that, it kills the trust that any teaching relationship depends on. Or perhaps it simply reminds us that that trust has always been a precious, much-abused thing. If I feel that my job now requires me to make judgments that are basically impossible—to tell an orderly, slightly stiff, reasonably good paper arrived at through hours of frustration from one arrived at through a minute's prompting and half an hour of light editing, for example—the job of my students has always been likewise impossible. There I am, demanding that they practice the extreme vulnerability of young adults learning in public, asking them to commit themselves to the study of things such as reading and writing that I consider to be living processes, open-ended and unmasterable. And there the surrounding society is—their justifiably anxious and perhaps indebted parents, who want them to be successful and happy; the corporate donors and partners that prestigious schools openly court and who want them to be productive and docile employees. What they want are people who have mastered various discrete and somewhat mechanized sets of skills. All of us insist on the life-and-death importance of a thing called 'education' while not remotely agreeing on what that thing is. And then there are the demands of their own big, half-formed, restless selves to consider too. What should we expect, but that they should take every shortcut in their doomed efforts to placate everyone? We asked them to work hard, but forgot to give them a consistent answer as to why. No wonder they cheat—they must already feel so cheated.
