January 2024

Mohammad Anas Wahaj | 14 jan 2024

According to Wikipedia, 'Generative artificial intelligence (generative AI, GAI, or GenAI) is artificial intelligence capable of generating text, images, or other media, using generative models. Generative AI models learn the patterns and structure of their input training data and then generate new data that has similar characteristics.' Positive aspects of generative AI systems include accelerative creativity, egalitarian tech for general public etc, while negative aspects include political propaganda through biased data, human resource displacement challenges etc. Commenting on transformative power of generative AI, Rich Palmer of Launchpad Venture Group, says, 'It's the new electricity.' Jonathan Griffiths, director of Babson College's Weissman Foundry, says, 'Much like a cell phone, AI is going to change how we interact with our computers and with each other in meaningful ways - and, if you don't have an understanding of what generative AI can do and what its limitations are, you're going to be left in the dust.' Joshua Herzig-Marx, a coach for early stage founders, says, 'At this point, if you have a startup and you don’t have a generative AI strategy, your board will be really unhappy with you, because that’s what everybody expects—in the same way that, if you didn't have a social strategy 15 years ago, it was a bad thing.' Prof. Ruth Gilleran and Prof. Clare Gillan of Babson College have designed a compulsory course for all undergraduates, 'Digital Technologies for Entrepreneurs'. Prof. Gillan says, 'We live in a time of tremendous disruption, and the pace of change has only accelerated. I want (students) to land on the right side of that continuous change.' Experts from Babson College provide insights and guidance on generative AI to entrepreneurs - (1) It will enable non-engineers to innovate in new ways: Prof. Gillaran says, 'It further democratizes the entrepreneurial process.' Prof. Thomas Davenport says, 'From an entrepreneurship standpoint, it lowers the barriers for tech expertise to design new products. It's a good thing for entrepreneurs.' (2) It should only be used in certain instances: Mr. Herzig-Marx says, 'Judgment is the big challenge (with generative AI), which is always one of the hardest things for any businessperson. There's no reason to think that whatever pops out of ChatGPT or a text-to-image service is going to be something you would actually want to use.' (3) Knowledge and content management will be transformed: Prof. Davenport says, 'Generative AI will rejuvenate the job of a knowledge manager...I think there are a lot of advantages to doing it for educating your frontline people and customer service applications.' (4) It will generate instant feedback, allowing entrepreneurs to assess viability quickly: Mr. Griffiths says, 'I could see (entrepreneurs) working with generative AI to solve the problems that they may not necessarily have the skills to solve right now.' Prof. Erik Noyes, who teaches Entrepreneurial Opportunities in AI, says, 'Generative AI enables the rapid prototyping of entrepreneurial ideas: literally a visualization and expression of an entrepreneurial idea that you can show to a target customer. You can get feedback on whether you're on a compelling path and creating value, or whether your idea is a dud.' (5) Beware of bias: Prof. Davit Khachatryan, who specializes in machine learning and data science, says, 'Generative AI is merely a means to an end, not an end in itself...Taking the results of generative AI at face value is like the blind following the blind. Today's entrepreneur, or any user of generative AI, needs to have an above-average understanding of how these tools work—and I think that’s where we analytics and data-science educators have a crucial role to play.' Prof. Noyes says, 'If the existing data is biased, there’s a strong likelihood that what’s generated can also be biased. You have to look at anything you’re doing in generative AI through the critical lens of 'How could this just be re-expressing bias?'' (6) Regulatory concerns could constrain creativity: Sam Altman, CEO of ChatGPT creator OpenAI, has urged international regulation of generative AI. Mr. Palmer says, 'When the front-runner (OpenAI) pushes for regulation, it opens up a question of whether anyone else can swim in the wake or not, and if anybody else can catch up again.' (7) Humans still matter: Prof. Khachatryan says, 'Overly relying on the seeming 'magic' that is provided by generative AI is not going to work. To have your leg up, you still need to put your creative hat on and keep it on at all times...it currently has no mechanism in place to evaluate the quality, meaningfulness, or effectiveness of these responses. I don’t think that one should get overexcited about how human-like the responses are because human-like, at the end of the day, doesn’t translate necessarily into meaningful.' Read on...

Babson Magazine: The Age of AI: Seven Things Entrepreneurs Need to Know
Author: Kara Baskin

Mohammad Anas Wahaj | 11 jan 2024

According to the research 'Reidentification Risk in Panel Data: Protecting for k-Anonymity' (Authors: Sachin Gupta of Cornell University; Shaobo Li of University of Kansas; Matthew J. Schneider of Drexel University; Yan Yu of University of Cincinnati), published on 07 oct 2022 in Information Systems Research, nearly all market research panel participants are at risk of becoming de-anonymized. The commitment of a market research company towards privacy of panelists cannot be totally practiced as there are ways around it. Prof. Sachin Gupta says, 'When organizations release or share data, they are complying with privacy regulations, which means that they’re suppressing or anonymizing personally identifiable information. And they think that they have now protected the privacy of the individuals that they’re sharing the data about. But that, in fact, may not be true, because data can always be linked with other data.' Earlier research (2006) 'How To Break Anonymity of the Netflix Prize Dataset' (Authors: Arvind Narayanan of Princeton University; Vitaly Shmatikov of Cornell University) showcases the similar risk. Researchers developed a de-anonymization algorithm, Scoreboard-RH, that was able to identify up to 99% of Netflix subscribers by using anonymized information from a 2006 competition, aimed at improving its recommendation service, coupled with publicly available info on Internet Movie Database. Both of these researchs rely on 'quasi-identifiers' or QIDs, which are attributes that are common in both an anonymized dataset and a publicly available dataset, which can be used to link them. The conventional measure of disclosure risk, termed unicity, is the proportion of individuals with unique QIDs in a given dataset; k-anonymity is a popular data privacy model aimed to protect against disclosure risk by reducing the degree of uniqueness of QIDs. Prof. Gupta suggests that even though privacy laws are getting tougher but market researchers will continue to collect and store data, and the challenge of privacy remains. He says, 'The nature of the problem will probably reduce and change, but I don't think it's going away. Read on...

Cornell Chronicle: Protecting identities of panelists in market research
Author: Tom Fleischman