Showing 1–8 of 8 results for author: Guadagni, G

Risks and Opportunities of Open-Source Generative AI

Authors: Francisco Eiras, Aleksandar Petrov, Bertie Vidgen, Christian Schroeder, Fabio Pizzati, Katherine Elkins, Supratik Mukhopadhyay, Adel Bibi, Aaron Purewal, Csaba Botos, Fabro Steibel, Fazel Keshtkar, Fazl Barez, Genevieve Smith, Gianluca Guadagni, Jon Chun, Jordi Cabot, Joseph Imperial, Juan Arturo Nolazco, Lori Landay, Matthew Jackson, Phillip H. S. Torr, Trevor Darrell, Yong Lee, Jakob Foerster

Abstract: Applications of Generative AI (Gen AI) are expected to revolutionize a number of different areas, ranging from science & medicine to education. The potential for these seismic changes has triggered a lively debate about the potential risks of the technology, and resulted in calls for tighter regulation, in particular from some of the major tech companies who are leading in AI development. This reg… ▽ More Applications of Generative AI (Gen AI) are expected to revolutionize a number of different areas, ranging from science & medicine to education. The potential for these seismic changes has triggered a lively debate about the potential risks of the technology, and resulted in calls for tighter regulation, in particular from some of the major tech companies who are leading in AI development. This regulation is likely to put at risk the budding field of open-source generative AI. Using a three-stage framework for Gen AI development (near, mid and long-term), we analyze the risks and opportunities of open-source generative AI models with similar capabilities to the ones currently available (near to mid-term) and with greater capabilities (long-term). We argue that, overall, the benefits of open-source Gen AI outweigh its risks. As such, we encourage the open sourcing of models, training and evaluation data, and provide a set of recommendations and best practices for managing risks associated with open-source generative AI. △ Less

Submitted 29 May, 2024; v1 submitted 14 May, 2024; originally announced May 2024.

Comments: Extension of arXiv:2404.17047

Near to Mid-term Risks and Opportunities of Open-Source Generative AI

Authors: Francisco Eiras, Aleksandar Petrov, Bertie Vidgen, Christian Schroeder de Witt, Fabio Pizzati, Katherine Elkins, Supratik Mukhopadhyay, Adel Bibi, Botos Csaba, Fabro Steibel, Fazl Barez, Genevieve Smith, Gianluca Guadagni, Jon Chun, Jordi Cabot, Joseph Marvin Imperial, Juan A. Nolazco-Flores, Lori Landay, Matthew Jackson, Paul Röttger, Philip H. S. Torr, Trevor Darrell, Yong Suk Lee, Jakob Foerster

Abstract: In the next few years, applications of Generative AI are expected to revolutionize a number of different areas, ranging from science & medicine to education. The potential for these seismic changes has triggered a lively debate about potential risks and resulted in calls for tighter regulation, in particular from some of the major tech companies who are leading in AI development. This regulation i… ▽ More In the next few years, applications of Generative AI are expected to revolutionize a number of different areas, ranging from science & medicine to education. The potential for these seismic changes has triggered a lively debate about potential risks and resulted in calls for tighter regulation, in particular from some of the major tech companies who are leading in AI development. This regulation is likely to put at risk the budding field of open-source Generative AI. We argue for the responsible open sourcing of generative AI models in the near and medium term. To set the stage, we first introduce an AI openness taxonomy system and apply it to 40 current large language models. We then outline differential benefits and risks of open versus closed source AI and present potential risk mitigation, ranging from best practices to calls for technical and scientific contributions. We hope that this report will add a much needed missing voice to the current public discourse on near to mid-term AI safety and other societal impact. △ Less

Submitted 24 May, 2024; v1 submitted 25 April, 2024; originally announced April 2024.

Comments: Accepted to ICML'24 as a position paper

Perturbation theory for a non-equilibrium stationary state of a one-dimensional stochastic wave equation

Authors: Gianluca Guadagni, Lawrence E. Thomas

Abstract: We address the problem of constructing a non-equilibrium stationary state for a one-dimensional stochastic Klein-Gordon wave equation with non-linearity, using perturbation theory. The linear theory is reviewed, but with the linear equations of motion including an additional potential term which emerges in the renormalization of the perturbation expansion for the state corresponding to the non-lin… ▽ More We address the problem of constructing a non-equilibrium stationary state for a one-dimensional stochastic Klein-Gordon wave equation with non-linearity, using perturbation theory. The linear theory is reviewed, but with the linear equations of motion including an additional potential term which emerges in the renormalization of the perturbation expansion for the state corresponding to the non-linear equations of motion. The potential is the solution to a fixed point equation. Low order terms in the expansion for the two-point function are determined. △ Less

Submitted 15 April, 2022; originally announced April 2022.

MSC Class: 82C31 (Primary) 70S20; 81T15 (Secondary)

Rigorous quantum field theory functional integrals over the p-adics I: anomalous dimensions

Authors: Abdelmalek Abdesselam, Ajay Chandra, Gianluca Guadagni

Abstract: In this article we provide the complete proof of the result announced in arXiv:1210.7717 about the construction of scale invariant non-Gaussian generalized stochastic processes over three dimensional p-adic space. The construction includes that of the associated squared field and our result shows this squared field has a dynamically generated anomalous dimension which rigorously confirms a predict… ▽ More In this article we provide the complete proof of the result announced in arXiv:1210.7717 about the construction of scale invariant non-Gaussian generalized stochastic processes over three dimensional p-adic space. The construction includes that of the associated squared field and our result shows this squared field has a dynamically generated anomalous dimension which rigorously confirms a prediction made more than forty years ago, in an essentially identical situation, by K. G. Wilson. We also prove a mild form of universality for the model under consideration. Our main innovation is that our rigourous renormalization group formalism allows for space dependent couplings. We derive the relationship between mixed correlations and the dynamical systems features of our extended renormalization group transformation at a nontrivial fixed point. The key to our control of the composite field is a partial linearization theorem which is an infinite-dimensional version of the Koenigs Theorem in holomorphic dynamics. This is akin to a nonperturbative construction of a nonlinear scaling field in the sense of F. J. Wegner infinitesimally near the critical surface. Our presentation is essentially self-contained and geared towards a wider audience. While primarily concerning the areas of probability and mathematical physics we believe this article will be of interest to researchers in dynamical systems theory, harmonic analysis and number theory. It can also be profitably read by graduate students in theoretical physics with a craving for mathematical precision while struggling to learn the renormalization group. △ Less

Submitted 24 February, 2013; originally announced February 2013.

Comments: 162 pages

Asymptotics for the Late Arrivals Problem

Authors: Carlo Lancia, Gianluca Guadagni, Sokol Ndreca, Benedetto Scoppola

Abstract: We study a discrete time queueing system where deterministic arrivals have i.i.d. exponential delays $ξ_{i}$. The standard deviation $σ$ of the delay is finite, but its value is much larger than the deterministic unit service time. We describe the model as a bivariate Markov chain, we prove that it is ergodic and then we focus on the unique joint equilibrium distribution. We write a functional equ… ▽ More We study a discrete time queueing system where deterministic arrivals have i.i.d. exponential delays $ξ_{i}$. The standard deviation $σ$ of the delay is finite, but its value is much larger than the deterministic unit service time. We describe the model as a bivariate Markov chain, we prove that it is ergodic and then we focus on the unique joint equilibrium distribution. We write a functional equation for the bivariate generating function, finding the solution of such equation on a subset of its set of definition. This solution allows us to prove that the equilibrium distribution of the Markov chain decays super-exponentially fast in the quarter plane. Finally, exploiting the latter result, we discuss the numerical computation of the stationary distribution, showing the effectiveness of a simple approximation scheme in a wide region of the parameters. The model, motivated by air and railway traffic, was proposed many decades ago by Kendall with the name of "late arrivals problem", but no solution has been found so far. △ Less

Submitted 29 December, 2016; v1 submitted 8 February, 2013; originally announced February 2013.

MSC Class: 60J10; 60K25

Rigorous quantum field theory functional integrals over the p-adics: research announcement

Authors: Abdelmalek Abdesselam, Ajay Chandra, Gianluca Guadagni

Abstract: In this short note we announce the construction of scale invariant non-Gaussian generalized stochastic processes over three dimensional p-adic space. The construction includes that of the associated squared field and our result shows this squared field has a dynamically generated anomalous dimension which rigorously confirms a prediction made more than forty years ago, in an essentially identical… ▽ More In this short note we announce the construction of scale invariant non-Gaussian generalized stochastic processes over three dimensional p-adic space. The construction includes that of the associated squared field and our result shows this squared field has a dynamically generated anomalous dimension which rigorously confirms a prediction made more than forty years ago, in an essentially identical situation, by Kenneth G. Wilson. We also prove a mild form of universality for the model under consideration. △ Less

Submitted 29 October, 2012; originally announced October 2012.

Queueing systems with pre-scheduled random arrivals

Authors: G. Guadagni, S. Ndreca, B. Scoppola

Abstract: We consider a point process $i+ξ_i$, where $i\in \bZ$ and the $ξ_{i}$'s are i.i.d. random variables with variance $σ^{2}$. This process, with a suitable rescaling of the distribution of $ξ_i$'s, converges to the Poisson process in total variation for large $σ$. We then study a simple queueing system with our process as arrival process, and we provide a complete analytical description of the syst… ▽ More We consider a point process $i+ξ_i$, where $i\in \bZ$ and the $ξ_{i}$'s are i.i.d. random variables with variance $σ^{2}$. This process, with a suitable rescaling of the distribution of $ξ_i$'s, converges to the Poisson process in total variation for large $σ$. We then study a simple queueing system with our process as arrival process, and we provide a complete analytical description of the system. Although the arrival process is very similar to the Poisson process, due to negative autocorrelation the resulting queue is very different from the Poisson case. We found interesting connections of this model with the statistical mechanics of Fermi particles. This model is motivated by air traffic systems. △ Less

Submitted 11 February, 2009; v1 submitted 29 May, 2008; originally announced May 2008.

Comments: 21 pages, 5 figures, new section on correlated arrivals, version 4

Report number: IMS-PS-PS_2008_131

Finite range Decomposition of Gaussian Processes

Authors: David C. Brydges, G. Guadagni, P. K. Mitter

Abstract: Let $\D$ be the finite difference Laplacian associated to the lattice $\bZ^{d}$. For dimension $d\ge 3$, $a\ge 0$ and $L$ a sufficiently large positive dyadic integer, we prove that the integral kernel of the resolvent $G^{a}:=(a-\D)^{-1}$ can be decomposed as an infinite sum of positive semi-definite functions $ V_{n} $ of finite range, $ V_{n} (x-y) = 0$ for $|x-y|\ge O(L)^{n}$. Equivalently,… ▽ More Let $\D$ be the finite difference Laplacian associated to the lattice $\bZ^{d}$. For dimension $d\ge 3$, $a\ge 0$ and $L$ a sufficiently large positive dyadic integer, we prove that the integral kernel of the resolvent $G^{a}:=(a-\D)^{-1}$ can be decomposed as an infinite sum of positive semi-definite functions $ V_{n} $ of finite range, $ V_{n} (x-y) = 0$ for $|x-y|\ge O(L)^{n}$. Equivalently, the Gaussian process on the lattice with covariance $G^{a}$ admits a decomposition into independent Gaussian processes with finite range covariances. For $a=0$, $ V_{n} $ has a limiting scaling form $L^{-n(d-2)}Γ_{c,\ast}{\bigl (\frac{x-y}{L^{n}}\bigr)}$ as $n\to \infty$. As a corollary, such decompositions also exist for fractional powers $(-\D)^{-α/2}$, $0<α\leq 2$. The results of this paper give an alternative to the block spin renormalization group on the lattice. △ Less

Submitted 28 August, 2003; v1 submitted 5 March, 2003; originally announced March 2003.

Comments: 26 pages, LaTeX, paper in honour of G.Jona-Lasinio.Typos corrected, corrections in section 5 and appendix A

MSC Class: 81T08;81T17;81T25;60G15;60G52;60G18